Collaborative Innovation Networks: Latest Insights from Social Innovation, Education, and Emerging Technologies Research [1st ed.] 978-3-030-17237-4;978-3-030-17238-1

Table of contents :
Front Matter ....Pages i-x
Front Matter ....Pages 1-1
Building a Shared Present and Future: Learnings from Henry Ford and Albert Kahn’s Co-Wuity Collaborative Innovation Network on the Moving Assembly Line and Mass Production (Ken Riopelle, Xin Wang)....Pages 3-41
Mirror, Mirror on the Wall, Who Is Leaving of Them All: Predictions for Employee Turnover with Gated Recurrent Neural Networks (Joao Marcos de Oliveira, Matthäus P. Zylka, Peter A. Gloor, Tushar Joshi)....Pages 43-59
Education and Technology as Levers for Sustainable Change (Carlo Alberto Amadei, Monica Baraldi Borgida)....Pages 61-72
Front Matter ....Pages 73-73
The Bezos-Gate: Exploring the Online Content of the Washington Post (Katharina Weitz, Florian Johren, Lukas Seifert, Sha Li, Jiexin Zhou, Oliver Posegga et al.)....Pages 75-90
Identifying Tribes on Twitter Through Shared Context (Peter A. Gloor, Andrea Fronzetti Colladon, Joao Marcos de Oliveira, Paola Rovelli, Manuel Galbier, Manfred Vogel)....Pages 91-111
Front Matter ....Pages 113-113
Social Media Teams of Hospitals as Mediators in Digital Health Ecosystems (Michael Beier, Sebastian Früh)....Pages 115-124
Promoting Holistic Care by Advancing Cultural Competence of Nursing Students in Mainland China (Hua Yuan, Caroline Porr)....Pages 125-131
Building Shared Environmental Governance for the Future: The Case of a Community COIN (Julia C. Gluesing, Ken Riopelle, Christina Wasson)....Pages 133-141
Effects of Innovation Efficiency and Knowledge on Industry-University Collaboration: An Evolutionary Game Perspective (Yang Song, Zhiyuan Zhang)....Pages 143-161
Front Matter ....Pages 163-163
Measuring Human-Animal Interaction with Smartwatches: An Initial Experiment (Katharina Stolz, Teresa Heyder, Peter A. Gloor, Oliver Posegga)....Pages 165-182
Show Me Your Moves: Analyzing Body Signals to Predict Creativity of Knowledge Workers (Marius Stein, Peter A. Gloor, Daniel Oster)....Pages 183-200
Promoting and Supporting Biodiversity Conservation Activities with the Pattern Language Approach: A Pattern Language for Collaborative Activities for Biodiversity Conservation (Arisa Kamada, Konomi Munakata, Mahito Kamada, Tomohiro Ichinose, Takashi Iba)....Pages 201-213
“Twelve-Tone Music Reloaded”: 12 Lessons in Rotating Leadership and Organizational Development from Jazz (Daniel C. Schmid, Peter A. Gloor)....Pages 215-225

Citation preview

Studies on Entrepreneurship, Structural Change and Industrial Dynamics

Yang Song Francesca Grippa Peter A. Gloor João Leitão Editors

Collaborative Innovation Networks Latest Insights from Social Innovation, Education, and Emerging Technologies Research

Studies on Entrepreneurship, Structural Change and Industrial Dynamics

Series editors João Leitão University of Beira Interior, Covilhã, Portugal Tessaleno Devezas University of Beira Interior, Covilhã, Portugal

The ‘Studies on Entrepreneurship, Structural Change and Industrial Dynamics’ series showcases exceptional scholarly work being developed on the still unexplored complex relationship between entrepreneurship, structural change and industrial dynamics, by addressing structural and technological determinants of the evolutionary pathway of innovative and entrepreneurial activity. The series invites proposals based on sound research methodologies and approaches to the above topics. Volumes in the series may include research monographs and edited/contributed works. More information about this series at http://www.springer.com/series/15330

Yang Song • Francesca Grippa • Peter A. Gloor • João Leitão Editors

Collaborative Innovation Networks Latest Insights from Social Innovation, Education, and Emerging Technologies Research

Editors Yang Song Jilin University Changchun, Jilin, China Peter A. Gloor MIT Center for Collective Intelligence Massachusetts Institute of Technology Cambridge, Massachusetts, USA

Francesca Grippa College of Professional Studies Northeastern University Boston, Massachusetts, USA João Leitão Department of Management and Economics University of Beira Interior Covilhã, Portugal

ISSN 2511-2023 ISSN 2511-2031 (electronic) Studies on Entrepreneurship, Structural Change and Industrial Dynamics ISBN 978-3-030-17237-4 ISBN 978-3-030-17238-1 (eBook) https://doi.org/10.1007/978-3-030-17238-1 © Springer Nature Switzerland AG 2019 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. The publisher, the authors, and the editors are safe to assume that the advice and information in this book are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or the editors give a warranty, express or implied, with respect to the material contained herein or for any errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. This Springer imprint is published by the registered company Springer Nature Switzerland AG. The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland

Preface

This book provides the latest research in the area of collaborative innovation networks (COINs) investigating how to best develop and deploy effective solutions to challenging and systemic business, social, and environmental issues. All 13 chapters included in this volume focus on innovative methodologies, technological platforms, and conceptual models to stimulate a collective discussion on how to leverage COINs to meet human needs in a sustainable way. Collaborative innovation networks are composed of small cyber-teams of intrinsically motivated individuals who are driven by a desire to find a solution to problems that affect their community and society in general. As such, they represent important building blocks of social innovation, as they provide the intellectual energy necessary for coping with external, unpredictable events. This book discusses the role of online social networks and technological networks in coping with uncertainty and decreasing excessive risks during highly stressful events. Part I will help you understand and frame the relationship between business and the environment adopting multiple disciplines and methodologies, including the archival case analysis method, design analysis, and neural network analysis. By reading this section, you will see how system analysis can be used to explore the positive changes that arise from collaborative efforts between business and the environment, in order to embrace environmental challenges and pursue reciprocal benefits. You will reflect on what makes entrepreneurs successful by reading about the fascinating partnership between Henry Ford and Albert Kahn, the Detroit architect who designed over 1000 buildings for Ford over more than three decades. And you will learn about the benefits of using deep learning to predict employee turnover through the analysis of e-mail communication networks over time. In Part II, you will learn how the application of sentiment analysis could shed light on the interdependency between mainstream media and business, by analyzing the content of newspaper articles based on the sentiment expressed toward specific companies. You will also explore a new system based on the longitudinal analysis of tweets to identify “tribes,” defined as networks of heterogeneous individuals connected by a shared passion or motivation. v

vi

Preface

In Part III, you will read interesting case studies of universities, government bodies, and hospitals across the world that optimized the role of collaborative innovation to advance cultural competence and mutual understanding. You will learn about the importance of leveraging social media and collaborative initiatives to improve collective decision-making when stakeholders from several disciplines are involved in complex scenarios. Lastly, Part IV explores novel approaches based on body sensors to measure and increase creativity and collaboration of knowledge workers in organizations. The next paper illustrates how to promote biodiversity conservation activity through design patterns. Another study applies body-sensing smartwatches to both horse and rider to improve human–animal interaction. The final case study will show you how a body-sensing system embedded into a smartwatch can allow individuals to predict their mood and help organizations create and maintain their “corporate groove” applying metaphors from Jazz. All four parts describe concrete and actionable applications of the power of collective decision-making and remind the reader of the importance of collaborative innovation when solving complex problems affecting society and the business world. Changchun, China Boston, USA Cambridge, USA Covilhã, Portugal

Yang Song Francesca Grippa Peter A. Gloor João Leitão

Acknowledgements

This volume combines selected contributions presented at the 8th International Conference on Collaborative Innovation Networks (COINs) held in Suzhou, China, from September 27 to 28, 2018. Scholars and practitioners discussed how to build a shared future through collaborative innovation networks across a variety of disciplines, including health, business, social innovation, entrepreneurship, and education. We would like to thank the sponsors of the conference who helped bring together international scholars, researchers, and practitioners to discuss how to create a shared future in large and small organizations, communities, and territories. These include China-Israel Center for Entrepreneurship and Innovation of Jilin University, China Center for Information Industry Development (CCID), Suzhou Economic & Information Technology Commission, China Future Industry 100, and the MIT Center for Collective Intelligence. This book is also supported by the National Natural Science Foundation of China (No. 71874068) and Youth Foundation of Humanities and Social Sciences, Ministry of Education of China (No. 17YJC790129). We are also grateful for the valuable insights provided by the Conference Program Committee members and to the many reviewers who provided precious feedback and constructive criticisms.

vii

Contents

Part I

Innovation Methods

Building a Shared Present and Future: Learnings from Henry Ford and Albert Kahn’s Co-Wuity Collaborative Innovation Network on the Moving Assembly Line and Mass Production . . . . . . . . . . . . . . . Ken Riopelle and Xin Wang Mirror, Mirror on the Wall, Who Is Leaving of Them All: Predictions for Employee Turnover with Gated Recurrent Neural Networks . . . . . . Joao Marcos de Oliveira, Matthäus P. Zylka, Peter A. Gloor, and Tushar Joshi Education and Technology as Levers for Sustainable Change . . . . . . . . Carlo Alberto Amadei and Monica Baraldi Borgida Part II

43

61

Innovation Applications

The Bezos-Gate: Exploring the Online Content of the Washington Post . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Katharina Weitz, Florian Johren, Lukas Seifert, Sha Li, Jiexin Zhou, Oliver Posegga, and Peter A. Gloor Identifying Tribes on Twitter Through Shared Context . . . . . . . . . . . . . Peter A. Gloor, Andrea Fronzetti Colladon, Joao Marcos de Oliveira, Paola Rovelli, Manuel Galbier, and Manfred Vogel Part III

3

75

91

Healthcare Applications

Social Media Teams of Hospitals as Mediators in Digital Health Ecosystems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115 Michael Beier and Sebastian Früh

ix

x

Contents

Promoting Holistic Care by Advancing Cultural Competence of Nursing Students in Mainland China . . . . . . . . . . . . . . . . . . . . . . . . . 125 Hua Yuan and Caroline Porr Building Shared Environmental Governance for the Future: The Case of a Community COIN . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133 Julia C. Gluesing, Ken Riopelle, and Christina Wasson Effects of Innovation Efficiency and Knowledge on Industry-University Collaboration: An Evolutionary Game Perspective . . . . . . . . . . . . . . . . . 143 Yang Song and Zhiyuan Zhang Part IV

Creativity

Measuring Human-Animal Interaction with Smartwatches: An Initial Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165 Katharina Stolz, Teresa Heyder, Peter A. Gloor, and Oliver Posegga Show Me Your Moves: Analyzing Body Signals to Predict Creativity of Knowledge Workers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183 Marius Stein, Peter A. Gloor, and Daniel Oster Promoting and Supporting Biodiversity Conservation Activities with the Pattern Language Approach: A Pattern Language for Collaborative Activities for Biodiversity Conservation . . . . . . . . . . . 201 Arisa Kamada, Konomi Munakata, Mahito Kamada, Tomohiro Ichinose, and Takashi Iba “Twelve-Tone Music Reloaded”: 12 Lessons in Rotating Leadership and Organizational Development from Jazz . . . . . . . . . . . . . . . . . . . . . . 215 Daniel C. Schmid and Peter A. Gloor

Part I

Innovation Methods

Building a Shared Present and Future: Learnings from Henry Ford and Albert Kahn’s Co-Wuity Collaborative Innovation Network on the Moving Assembly Line and Mass Production Ken Riopelle and Xin Wang

Abstract The purpose of this paper is to demonstrate how to build a shared future through Collaborative Innovation Networks (COINs) using a historical case study of the timely and unique Co-Wuity partnership between Henry Ford, the automotive industrialist, and Albert Kahn, the Detroit architect who designed and built over 1000 buildings for Ford over 34 years from 1908 to 1942. Wuity is a higher cognition with the flexible and imaginative ability to gain intuitive understanding of underlying phenomena and connect deliberative thinking for insightful problem-solving through a process of mindful observation and visual analogy. Ford was building an automotive manufacturing company, which was designing and manufacturing a vehicle in standardized high volumes and that needed a new building to meet that growing demand. Kahn was building an architectural firm, which was designing custom building plans on a project-by-project basis to meet the needs of his clients, also on a high-volume scale. Both developed Wuity behaviors of mindful observation and visual thinking to gain insights for innovations, experimentation, and implementation to see and understand what worked in their respective ventures, influencing each other’s businesses in the process. The focus of the case is on the emerging and future technologies at that time, on how COINs strengthened adaptability and transformability to leverage novel ideas as a competitive advantage using the two early nineteenth-century start-ups: the Ford Motor Company and Albert Kahn Architects and Engineers as our historical case study.

K. Riopelle (*) Department of Industrial and Systems Engineering, Wayne State University, Detroit, MI, USA X. Wang School of Humanities and Social Science, Beijing Institute of Technology, Beijing, China e-mail: [email protected] © Springer Nature Switzerland AG 2019 Y. Song et al. (eds.), Collaborative Innovation Networks, Studies on Entrepreneurship, Structural Change and Industrial Dynamics, https://doi.org/10.1007/978-3-030-17238-1_1

3

4

K. Riopelle and X. Wang

Our goal is to inspire today’s leaders by demonstrating the lessons these past leaders taught us about how to build a shared future through a Collaborative Innovation Network. In addition, this case extends the analysis of Wuity to a Co-Wuity Collaborative Innovation Network.

1 Introduction The year was 1896, and two entrepreneurs had launched their careers in the City of Detroit, Michigan, which would not overlap until 12 years later in 1908 and change the course of history for the future of industrial mass production and factory architectural design. In January 1896 Albert Kahn, age 27, launched a new architectural firm with two partners. Six months later in June 1896, Henry Ford, age 33, took his first gasolinepowered quadricycle for a drive through the City of Detroit. Thus, begins the historical case study of the circumstances and context that would result in a 34-year collaborative learning experience from 1908 to 1942 and the design and construction of some 1000 buildings for these two aspiring young men. The paper is organized into the following sections: • Literature Review with a brief review of mindfulness and Wuity research. • Background which enumerates significant events in the State of Michigan at the time; Similarities and Differences between Henry Ford and Albert Kahn; and The Intersection and Social Interaction of Henry Ford and Albert Kahn. • Factory Plant: Three Case Studies, which include the Highland Park Plant, the Rouge Plant, and the Willow Run Bomber Plant, this section describes how Albert Kahn helped Henry Ford. • Mindful Social Interaction and Shared Present and Future: this section describes how Henry Ford helped Albert Kahn. • Lessons Learned for Today’s Entrepreneurs. • Conclusion and Future Research. The Background section will provide an orientation to the historical times at the turn of the twentieth century and the key fundamental resources that the two future leaders would draw upon; the Factory Plant: Three Case Studies section will describe the developmental sequence that Ford and Kahn’s partnership went through from the 1908 Highland Park Plant with the launch the Model T, to the 1915 colossal Rouge Plant Complex, and to the 1941 Willow Run Bomber Plant; the Lessons Learned section summarizes what we have learned about Co-Wuity and COINs with suggestions for today’s entrepreneurs; and the paper ends with our conclusions and suggestions for future research. Overall, the paper builds the picture of the social interaction between these two leaders and their respective teams and how each learned from the other to achieve their business goals. In short, this is a story of the Co-Wuity Collaborative Innovation Network.

Building a Shared Present and Future: Learnings from Henry Ford and. . .

5

2 Literature Review Mindfulness is an important topic in cognitive science, medical science, psychology, and management. After 40 years of study, mindfulness has been recognized as a state of mind having a wide conscious attention and an altered trait such as neuroplasticity from long-term practice, which creates both structural and functional brain changes (Dane 2011; Goleman and Davidson 2017; Langer 1989). Weick and Sutcliffe in their three books Managing the Unexpected use the term “mindfulness” as a higher-order category for five specific behaviors that are found in high reliability organizations (Weick and Sutcliffe 2001, 2011, 2015). They illustrate the presence of these behaviors in exemplary stories from nuclear power plants, firefighting crews, hospitals, banks, and more. Rather than a focus on how to avoid failure or quickly recover from a failure, the emphasis in our paper is on processes of creativity, innovation, and building something new. Wuity is a higher cognition with the flexible and imaginative ability to gain intuitive understanding of underlying phenomena and connect deliberative thinking for insightful problem-solving through a process of mindful observation and visual analogy. Wuity is a collective thinking model long used for centuries in China, but it is also an inherent higher cognitive skill through training the mind by being mindful in modern society. Wang and her coauthors introduced Wuity to management innovation (Wang and Gloor 2018; Wang and Li 2017, 2018). For innovators, scientists, and entrepreneurs, a series of Wuity six steps have been identified to help gain intuitive understanding, insights, and innovation. They are getting stuck, letting it go, mindful observation, visual analogy, gaining insights, and implementation (Wang and Gloor 2018). “Getting stuck” means an individual or team has come to an impasse and cannot solve a problem or issue. “Letting go” means taking a step back and abandoning the current train of thinking, discarding preconceived ideas. “Mindful observation” is observing and seeing with awareness, which is the careful attending to internal and external phenomena (e.g., thoughts, emotions, sounds, smells, or proprioceptive sensations). Visual analogy is an analogical reasoning process based on the similarity, connection, and common essence among dynamic perceptual imagery and mental images. Visual mental images have led many people to liken this concept to “seeing with the mind’s eye.” “Gaining insights” means to understand how different disciplines and domains are related to one another at the most basic level. Gaining a clear and deep insight is the most productive and transcendental result of Wuity thinking. Once a good analogy between different domains is found, confusion and contradiction caused by superficial observations disappear, and the result is clarity and meaningful insights for problem-solving. “Implementing” means the ability to put into practice those insights gained from the process of letting go, mindful observation, and visual analogy. Previous Wuity research has focused on the level of individual and intrapersonal, whereas in this paper, the focus is on the interpersonal level or between people, especially from a social network view. Gloor in his recent books (Gloor 2017a, b) identifies six directional indicators for high-performing teams based on over 14 years of studying Collaborative Innovation

6

K. Riopelle and X. Wang

Networks or COINs. These studies have relied on the use of social network analysis of email, blogs, web sites, and twitter data to identify and confirm the credibility of these indicators on team performance. Since our paper is a historical case study of a relationship that occurred over 122 years ago, we rely upon archival books and other sources to reconstruct our case. Gloor’s six indicators are central leadership, rotating leadership, balanced contribution, rapid response, honest language, and shared context.1 For the purposes of this paper, we will provide a nontechnical definition of these six indicators. “Central leadership” means a clear visible leader who acts as an information hub and controls information and decision-making. “Rotating leadership” means other team members may take turns and assume a leadership position besides the central leader. “Balanced contribution” means team members equally listen, direct, and contribute with no member always dominating or being silent. “Rapid response” means team members are quick to respond to other team member requests, questions, or inquires. “Honest language” means team members are truthful and use both positive and negative language in their communication. “Shared context” means that team members use a common vocabulary as well as understand the same fundamental norms for how work gets done and how to relate to one another. Our contribution to the literature is an interweaving of two streams of research based on the six process steps of Wang’s Wuity and Gloor’s six directional indicators of high-performance teams using an historical archival case analysis method as exemplified by Weick and Sutcliffe.

3 Background 3.1

State of Michigan

In 1805 Detroit burned to the ground and was destroyed. Later Detroit would declare its city motto to be “We hope for better days; it shall rise from the ashes” [Detroit Historical Society Timeline (1787–1820, 1820–1860, 1860–1900)]. Over the next 90 years, not only did Detroit rise from the ashes, but the city was to become a thriving, growing, diverse, immigrant community. Detroit was being networked to the East, Midwest, North, and South and beyond by an intermix of infrastructures including rail, shipping waterway canals, stagecoach, and mail networks. New technologies were being developed and experimented with including steel, horseless carriages along with the installation of a new communications media of the telephone. Newspapers were being developed, and institutions of higher education were founded. In 1835 the population of Michigan would hit the threshold of 85,000 to make it the 26th state of the Union. The US Civil War was being fought from 1861 to

1 For the technical network definitions and calculations of Gloor’s six directional indicators of highperformance teams (Gloor 2017a, see Table 3, pp. 52–53).

Building a Shared Present and Future: Learnings from Henry Ford and. . .

7

1865 when Abraham Lincoln was president, and during that time Henry Ford was born. Above all else, Detroit became a hub for talented risk-taking entrepreneurs to launch their dreams. This concentration and growth of physical infrastructures, new communication technology, new industries, and the influx of immigrants—Germans, Scandinavians, and others—created the population boom that seeded the fertile and rich ground for Henry Ford and Albert Kahn to meet and to not only grow but thrive during their Co-Wuity relationship. Table 1 lists over 25 significant events that chronicle many of these historical times. Specifically, the chronology of events in this case begins with the founding of the University of Michigan (UofM) in 1817, an institution of higher education that would become a multithreaded resource in the life of Albert Kahn. UofM was the institution of higher education where three of his five younger brothers (Julius, Moritz, and Louis) would receive their degrees in civil engineering and architecture and join him in his firm; his future wife, Ernestine, would graduate from UofM with an Arts & Letters degree just before they were married; and UofM itself would become a major client contracting some 26 major buildings (Matuz 2002). The shipping waterway was expanded with the completion of the Erie Canal (1825) and the Sault Ste. Marie locks (1855), which opened iron ore and copper mining in Michigan’s Upper Peninsula and fueling the rise of Detroit becoming a major shipping port in 1850 and supporting the North’s Civil Wars offense of the manufacturing of weapons between 1861 and 1865 and connecting Michigan and Detroit to the east coast. Rail transportation expanded and grew from the completion of the DetroitPontiac Railway (1838), the Detroit and New York City link (1854), and the building of railroad cars in 1853. And in 1896, the last horse-drawn street cars were replaced with electric trolleys. Mail and road construction grew with mail service established in 1831 between Detroit and the east coast, as well as daily stagecoach service going west between Detroit and Chicago in 1836. New technologies and new companies were being founded including the first Bessemer-type steel manufacturing operation (1864) laying the groundwork for railroad, stove, and automobile manufacturing in Detroit. Detroit became the first city to assign individual telephone numbers (1879), and in 1895 Detroit’s municipal power station begins to supply power for streetlights and public buildings. The D. M. Ferry Seed Company is founded in 1867 and would become the largest in the country; this seed supply was an important source for urban families to grow their small plots of vegetables and landscape their homes and gardens with flowers, which Albert Kahn and his family found of great value upon his arrival in Detroit in 1891 and later in landscaping his home and summer residence. The German wave of immigration in 1825 along with the Scandinavian influx of 1849 brought skilled workers and artisans to Detroit. Specifically, in 1849 Julius Melchers, a German artist and sculpture, arrived in Detroit and would become an early important mentor for Albert Kahn.

8

K. Riopelle and X. Wang

Table 1 Detroit historical events 1817 1825

1827

1831 1836 1837

1838 1845 1849 1850 1853 1854 1855 1861–1865 1863 1864 1867 1869 1870s– 1885 1879 1881 1890 1895 1896

The University of Michigan (UofM) is founded and first classes held in 1841 October 25. The Erie Canal is completed. It connects Lake Erie with the Hudson River near Albany, New York, making transportation of people and goods across New York State easier, faster, and less expensive. Roads are improved between Detroit and Toledo, St. Joseph, and Chicago The first wave of German immigration to Detroit begins Detroit adopts its city seal and motto: Speramus Meliora; Resurget Cineribus. It means “We hope for better days; it shall rise from the ashes.” It commemorates the Fire of 1805 Mail service begins between Detroit and the east coast Daily stagecoaches begin running between Detroit and Chicago The Michigan Territory’s population reaches 85,000, which qualifies it for statehood. Michigan becomes the 26th state of the United States of America. Detroit is its first capital The Detroit-Pontiac railway is complete. The first cars are horse-drawn Iron ore mining begins in the Upper Peninsula Scandinavian immigrants arrive in Michigan. Many settled in Michigan’s Upper Peninsula, but some work in Detroit as toolmakers Shipping becomes Detroit’s biggest industry George B. Russell begins building railroad cars under the business name Detroit Car and Manufacturing Co The first rail connection between Detroit and New York City is completed The locks at Sault Ste. Marie are finished. They enable ships to bring iron and copper ore to Detroit entirely by ship US Civil War. Copper mined in Michigan for manufacturing of weapons (Matuz 2002) July 30, Henry Ford is born The first Bessemer-type steel is produced at Eureka Iron Works in Wyandotte, laying the groundwork for railroad, stove, and automobile manufacturing in Detroit D. M. Ferry Seed Company is founded March 21, Albert Kahn is born in Rhaunen, Kingdom of Prussia, which is now Germany Karl Benz in Mannheim, Germany, and Gottlieb Daimler in Cologne, Germany, each develop an internal combustion engine and motor car (Hildebrand 1974) Detroit becomes the first city to assign individual telephone numbers Detroit is the center of the nation’s stove manufacturing industry Albert Kahn and family arrive in Detroit (Matuz 2002) Detroit’s population reaches 205,876. It ranks 15th in size among American cities Detroit’s municipal power station begins to supply power for streetlights and public buildings Albert Kahn forms Architectural Company March 6. Charles Brady King drives Detroit’s first gasoline-powered car down Woodward Avenue June 4. Henry Ford test drives his first car, using engine parts borrowed from Charles Brady King The last horse-drawn street cars are replaced by electric trolleys

Events from Detroit Historical Society website Timeline of Detroit. https://detroithistorical.org/ learn/timeline-detroit/ Last accessed November 12, 2018

Building a Shared Present and Future: Learnings from Henry Ford and. . .

9

Detroit’s population balloons from under 1000 in 1805 to 205,876 in 1890 ranking it as the 15th largest American city. This background of people, places, and events was the conditional context surrounding the lives of Henry Ford and Albert Kahn before their business relationship and co-learning occurred. Context is “a way of life and work in a specific geographical area with its own set of business conditions, cultural assumptions, and unique history” (Gluesing et al. 2003, p. 354). To complete the background, the next section is a comparison of the similarities and differences between these two men. The comparison demonstrates that there was a great deal of common ground between Henry Ford and Albert Kahn that served to speed the formation of a trusted relationship.

3.2

Similarities Between Henry Ford and Albert Kahn

Henry Ford and Albert Kahn had eight distinguishing similarities including a shared geographical location, birth order, education, childhood hardships, significant mentorships, early business hardships, and a preference for and reliance on talented teams to perform work, and both were visual thinkers. This shared context and common life experience enabled a cognitive convergence and supported shared values for hard work and ways of working that ultimately lead to a trusted relationship that each could depend upon. A short description of the seven similarities follows. 1. Both were born in the 1860s and grew up in the Detroit area. Henry was born in 1863, in Dearborn, a suburb of Detroit. Albert was born 6 years after Ford in 1869 and moved to Detroit at age 12, in 1891. 2. Both were the firstborn sons within their families. Albert had five younger brothers and two sisters; Henry had two younger brothers and two sisters. 3. Both were self-educated. Ford had a formal education from age 7 to 16, but began his self-education by tinkering and repairing watches at age 13 (Curcio 2013). Albert’s formal education ended around age 11, when he left Europe. Albert’s self-education began when drawing and sketching before the age of age 11 (Matuz 2002). 4. Both had early hardships growing up. Ford’s mother died when he was 13 (Curcio 2013). Albert assumes male head of household at age 12, while his father travels (Matuz 2002). 5. Both had significant mentors. Ford’s friend, Charles B. King, credited with the first quadricycle, lent Ford parts for his quadricycle [INDUSTRIAL DETROIT (1860–1900)].

10

K. Riopelle and X. Wang

And, Thomas Edison encouraged Ford and employed him at the Detroit Edison Illuminating Company. Khan was mentored by family friend, sculpture Julius Melchers; a local prominent architect, George Mason, at the firm Mason and Rice Architects; and a contemporary architect, Henry Bacon, during his and Kahn’s European scholarship tour in 1901 (Matuz 2002). 6. Both had difficult and short-lived early business partnerships. Ford’s first two companies failed before starting the Ford Motor Company in 1903. Kahn’s first company, Nettleton, Kahn, and Trowbridge in 1896, had a rocky start because in 1897 George Nettleton left to take a university professorship at Cornell and Alexander Trowbridge died in 1900. 7. Both relied on talented teams that worked for them. Ford in the early days had James Couzens, C. Harold Wills, Walter Flanders, and Charles Sorensen who are credited with contributing to the early success for the Model T (Sorensen et al. 2006). Kahn employed and was supported by his three younger brothers, Julius, Moritz, and Louis, along with talented architects, such as Ernest Wilby and Wirt Rowland (Smith 2017). Both Ford and Kahn learned how to organizationally execute the timely flow of materials through their respective organizations to achieve a high volume of quality output in a cost-effective manner. 8. Both were visual thinkers. Kahn from his early childhood drawing and later as an architect perfected his visual acuity. Ford used models to visually assess their value and utility. In summary, the shared similarities provided the basis for a long, respectful, and trusting relationship. Still, there were some noteworthy differences.

3.3

Differences Between Henry Ford and Albert Kahn

1. Family support for career choices differed. Henry Ford’s father wanted him to be a farmer like himself; however, Henry Ford disliked farming and had no inclination to take up that profession and lifestyle. Kahn’s mother, father, aunt, and close family friends supported his early music and drawing talent, by providing piano lessons, drawing lessons, and equipment, plus job referrals to launch his architectural career. 2. Personalities were different. Ford was a visionary introverted industrialist always looking ahead to the next phase of industrial production but, just the same, relished in the spotlight. Ford rarely entertained guests at his home in Fair Lane (Sorensen et al. 2006).

Building a Shared Present and Future: Learnings from Henry Ford and. . .

11

Kahn was a talented, nontraditional architect extrovert, who worked tirelessly to solve the practical architectural problems with design and building innovations. He was humble and always credited his client, especially Henry Ford, for insights. In contrast to Ford’s preference for personal solitude, Kahn was either entertaining or being entertained and had no less than 20 guests regularly for Sunday brunch at his summer house (Hodges 2018). 3. Siblings’ involvement in their businesses was different. Ford’s two brothers and two sisters are never mentioned in his building of the Ford Motor Company. However, Ford’s only son, Edsel, did join the company, as well as his grandson, Henry Ford II. Kahn’s three brothers, Julius, Moritz, and Louis, would all join his company and make significant contributions both technically and organizationally. Overall, the differences were few and non-consequential for the Ford and Kahn business relationship. Each trusted the other and knew they could depend upon each other’s word and organization to fulfill their respective commitments.

3.4

The Intersection and Social Interaction of Henry Ford and Albert Kahn

Henry Ford and Albert Kahn came together for primary two reasons: first, the nature of their work completed each other and second, that each valued speed of response. Ford was building an automotive manufacturing company, which was designing and manufacturing an automotive vehicle in standardized high volumes, and he needed a new building to meet demand. Kahn was building an architectural firm, which was designing building plans on a project-by-project custom basis to meet the needs of his clients, also on a high-volume scale. Both developed Wuity behaviors of mindful observation or visual thinking to gain insights for innovation, experimentation, and implementation to see and understand what was working in their respective business ventures and influenced each other’s businesses. They shared learning from their experiences with one another which resulted in personal and business gain for both men. Both Ford and Kahn valued speed. For Ford, speed meant increasing production, delivery, sales, and profits. Slowdowns along the logistic supply line meant a stop in production and a loss of sales and profits. For Kahn speed meant satisfied clients and an increase in the number of contracts the firm could process and a way to build the business reputation to generate even more business. In summary, Ford needed an architect to build what he wanted, and Kahn could deliver it with speed and on time completion and without extra costs, and he always made sure to give Ford credit for his innovation and insights. The next section is a description of the three historic buildings that defined and exemplify Ford’s and Kahn’s relationship.

12

K. Riopelle and X. Wang

4 Factory Plant: Three Case Studies Ford and Kahn’s partnership went through a developmental sequence from the 1908 Highland Park Plant with the launch of the Model T, to the 1915 colossal Rouge Plant Complex, and finally to the 1941 Willow Run Bomber Plant. The purpose here is to articulate the Co-Wuity creation process between Ford and Kahn and the social interaction that resulted in their mutual business success. The primary question of this paper is how do Wuity and high-performance teams relate to each other? We are interested in the lessons learned from this historical relationship between Henry Ford and Albert Kahn, two entrepreneurs at the beginning of the twentieth century, that can serve as a guide for today’s entrepreneurs who are innovating and building a shared future. Figure 1 presents the Wuity six process steps as a spiral diagram, which illustrates that the process can repeat itself but begins at a new starting point for continued growth and expansion. Table 2 summarizes and juxtaposes Wuity’s six process steps and Gloor’s six indicators of high-performance teams. The following three factory plant case studies

Fig. 1 Wuity six process steps spiral [Reproduced with permission from (Wang and Gloor 2018) p. 173]

Building a Shared Present and Future: Learnings from Henry Ford and. . .

13

Table 2 Wuity’s six process steps (Wang and Gloor 2018, pp. 172–173) and Gloor’s six indicators of high-performance teams (Gloor 2017a, p. 55) Wuity six process steps 1. Getting stuck 2. Letting go 3. Mindful observation 4. Visual analogy 5. Gaining insights 6. Implementing

Gloor’s six indicators of high-performance teams 1. Central leadership 2. Rotating leadership 3. Balanced contribution 4. Rapid response 5. Honest language 6. Shared context

Note: Brackets “[]” will be used within in the following text to indicate where we have identified Wuity six steps and Gloor’s six high-performing team indicators

will illustrate how these Wuity process steps and team indicators were interwoven to build a successful future for Henry Ford and Albert Kahn.

4.1

The Highland Park Plant

In 1908 Henry Ford [central leader] recognized that his current factory for automobile manufacturing could not keep up with the demand for automobile orders, which was driven by the “bulldog driving energy” of James Couzens, his Piquette Plant manager. Couzens’s constant nagging of dealers and branch agencies to buy and sell represents one team member’s [balanced contribution] and [honest language] with the constant message of we need more space! (Sorensen et al. 2006, p. 84). Other team members who shared the same plant space [common context,] could see for themselves that Couzens was right. The Ford Piquette Plant was a typical masonry construction building (Fig. 2), with floors supported by wooden beams and exterior load-bearing walls (Smith 2017, Chap. 1, Working for the Dean. Fig. 6). Ford himself realized that Couzens was right and bought an initial 60 acres of land in Highland Park as the site of his new factory. He contacted Albert Kahn, a local Detroit architect, to build a new factory there. Figure 3 is a picture of the interior of the Ford Piquette Plant showing the many interior columns and the fire wall door at the back. Ford was [getting stuck] to solve the problem of the slow expensive automobile production. He and his team members [let go] of the prevailing idea of moving the worker to the work and had the [mindful observation] of moving the work to the worker through [visual analogy] in observing how meat packers and mill houses worked, with the [gaining insight] of a moving assembly line as a new idea to be [implemented] in this new plant. The following provides some context for the development of the assembly line (Nevins and Hill 1954).

14

K. Riopelle and X. Wang

Fig. 2 Ford Piquette Plant (For more photos of the Ford Piquette, see https://www.google.com/ search?q¼Pic:Ford+Piquette+Plant&rlz¼1C1CHBF_enUS739US739&tbm¼isch&tbo¼u& source¼univ&sa¼X&ved¼0ahUKEwihzuPbvJnbAhWRFXwKHfNxCg4QsAQIfg&biw¼1833& bih¼1196 Last accessed on November 30, 2018) https://en.wikipedia.org/wiki/Ford_Piquette_ Avenue_Plant. (CC BY-SA 4.0)

Fig. 3 Ford Piquette Plant interior source: http://www.oldcarsweekly.com/news/henry-fordssecret-experimental-room-open-ford-piquette-plant, September 20, 2017

Building a Shared Present and Future: Learnings from Henry Ford and. . .

15

The continuous conveyor belts to bring materials up to assembly lines were the work primarily of Ford and Sorensen. A mechanic named Gregory, who had once worked in a brewery which used conveyors to lift grain from storage bins to mash tanks, suggested late in 1912 that a line of moving hoppers be employed to carry mixed core sand to chutes above the mould-makes’ [sic] bench. Of the group interested in the development of moving assembly lines, Clarence W. Avery had the broadest grasp of the subject and showed the most intelligent initiative. Early in 1912 Avery joined the Ford Motor Company, whereas Sorensen’s assistant gained a rapid knowledge of machine tools and technological processes [shared context]. To be sure, Ford [central leader] took a special interest in the magneto assembly, but elsewhere able employees like Gregory, Klann, and Purdy made important suggestions [balanced contribution]; Sorensen [rotating leadership] and others then helped work them out, while Ford [central leadership] gave encouragement and counsel. The largest single role in developing the new system, however, was played by this university-trained thinker, Avery, so recently brought in from his schoolroom. (pp. 472, 474)

Ford wanted Albert Kahn to build his Highland Park Plant because Kahn had generated a reputation for building an innovative automotive plant for a competitor, James Joy, manager of the Packard Motor Car Company. Ford was especially impressed with Packard Building 10 that was built in Detroit in 1906. Kahn’s design of Building 10 had solved some of the fundamental problems in previous designs. The following passage describes the problems that were present in previous factory designs and how Kahn approached and solved them (Matuz 2002). Factory buildings at the time were cramped and required thick walls and many columns running from the floor to the ceiling to support the floors and roof above. Parts of the floors, walls, and columns were often made of wood, and they easily became fire hazards when oil and greased soaked in. (p. 49)

Albert was not satisfied with his progress with Joy’s ideas using the Packard old style factory design in the first nine buildings. Albert had become [stuck]. However, for the tenth building, he imagined a new modern factory design [letting go] (ibid). As he began preparing the tenth building, he approached the design like a problem that needs to be solved. The problem is [mindful inquiry]: how can he create more inside space by having fewer support columns? “What we need is a large, open space, where workers will be comfortable,” [mindful observation and honest language]. Albert explained to Julius, his civil engineer brother, “We need to support the weight of the building without dozens of bulky columns running from the ceiling to the floor [visual analogy]. We need to provide enough space for people to work in a single area. Also, we also need to find ways to let sunlight and fresh air in.” [mindful observation and honest language] (p. 49)

The new idea was to design and build a concrete structure with steel reinforcement using his brother [gaining insights], Julius’s patented Kahn Bar [rotating leadership and shared context], who formed his own company, the Trussed Concrete Steel Company holding this valuable patent in 1902 (Matuz 2002, p. 47). The idea of using steel-reinforced concrete grew out of Albert’s and Julius’s [shared context] of building the University of Michigan’s Engineering Building in 1903 and the Pierce Great Arrow Automobile Plant in Buffalo, NY, in 1906 (Bucci 2002, pp. 33, 37).

16

K. Riopelle and X. Wang

Thus, Ford could see that the Packard Building 102 solved the obstruction issue caused by numerous interior columns and walls. This open factory design created a large free space for manufacturing flexibility. Plus, it had the added advantage of lots of sunlight from the large wall side windows, which could be open for ventilation. The construction had a further benefit of being more fireproof than a wood-supported structure, which reduced insurance costs as well [implementation]. However, Ford’s vision was bigger than just a building. The story unfolded in 1908 when Ford called Kahn to design Highland Park Plant and asked abruptly (Op cit.): “Mr. Kahn, can you design more efficient factories?” [central leadership] “Mr. Ford, I can design anything,” [rotating leadership] They met a week later at the Ford Factory on Piquette Ave. in Detroit. Then Ford drove them to the proposed Highland Park Plant 230-acre site: “I want to make a car that is inexpensive, easy to use, and produced in large quantities.” . . . “I plan to have the entire car built under one roof, from start to finish.” [honest language] “Impossible,” thought Albert . . . but he was still intrigued. [honest language] Then returned to Piquette office where Kahn made a sketch putting the building in the middle of the property. Ford, interrupted, “You’ve only got part of the idea. . . . I want the first building constructed right up along Woodward Avenue so that I have room to expand behind, to other parties [sic]of the property. This factory will take us into the future. Everything will be geared to building cars faster. Each worker will perform only one task, with all the parts and tools they need right by their side. Someday, instead of having workers move around to stations, we will have a moving assembly line that bring work to the workers.” [shared context] (p. 52)

Figures 4 through 8 represent five different images of the Packard Building 10. Figure 4 is a drawing of the exterior of the original two-story building measuring 60  322 feet with 30-foot column spans. An additional two stories were added later as shown in Fig. 7. Figure 5 is a detailed sectional view of the Kahn system of reinforced concrete illustrating the bent steel bars which add strength for the weight bearing columns. Figure 6 is an aerial view of Packard Plant Complex from 1903 to 1910. Figure 7 is an exterior view of Building 10 showing the columns and glass windows, now as a four-story building. Figure 8 is an interior photo of the large free space between the columns and the wall windows that allowed abundant sunlight to fill the factory floors. From this exchange we can see Henry Ford in his role as a [central leader] using [honest language] and a [shared context] to solve a [getting stuck] problem, the slow expensive automobile production. Ford wanted a [rapid response] from Kahn, and he got it. Kahn made an immediate sketch and perfected it in a back and forth collaboration with Ford. In less than 18 months, Ford’s Highland Park Plant,3 four stories high (85  860 feet), opened on New Year’s Day 1910 (Matuz 2002).

2

For photos of the Packard Building #10, see http://ilovedetroitmichigan.com/detroit-architecture/ albert-kahn-400-buildings-in-metro-detroit/ #63 to #68 Packard Building. Last accessed on May 22, 2018. 3 For photos of the Ford Highland Park Plant, see http://ilovedetroitmichigan.com/detroit-architec ture/albert-kahn-400-buildings-in-metro-detroit/2/ #104, #105, #106, #107 and #108. Last accessed on May 22, 2018.

Building a Shared Present and Future: Learnings from Henry Ford and. . .

17

Fig. 4 Packard Motor Car company, building 10, Detroit Michigan. Drawing (Bucci 2002, p. 30), public domain

Fig. 5 Kahn system of reinforced concrete detail (Bucci 2002, p. 30) public domain

18

K. Riopelle and X. Wang

Fig. 6 Packard Motor Car Company 1903–1910, Detroit Michigan. Aerial view. (Bucci 2002, p. 32) public domain It was nicked name the “Crystal Palace”, because its windows which ran along the side of every floor and parts of the roof, glistened in the sunlight and provided great light for work going on inside. The building’s large open space allowed all production to take place under one roof. . . . It allowed for assembly of a car to progress to different areas, where different parts were stored. That would be impossible in the old mills, where many columns interfered with free movement. (p. 55)

The four Figs. 9, 10, 11, and 12 depict Ford Motor Company’s Highland Park Plant Complex. Figure 9 is the Highland Park four-story sketch (85  860 feet) with 20-foot spans and no interior dividing walls. Compared to the Packard Building 10, the Highland Park Plant is 42% or 25 feet wider, 167% or 538 feet longer, and twice the height at four stories. Figure 10 is an aerial view of the Highland Park Plant Complex in 1915. Several buildings were added after the initial plant building was built, including an administrative building and a power house featuring five smokestacks. Figure 11 is an exterior view of the Highland Park Plant. Notice the corner extension and other vertical extensions along the long side of the building as seen in Fig. 9; these building extensions housed the auxiliary facilities such as changing rooms, bathrooms, and hydraulic freight elevators which transported material between the floors. Figure 12 is the interior view showing the four floors and the cavernous space beneath a roof of glass or monitors. Completed work of part subassemblies would be placed in bins and lowered down to the next floor until reaching the ground level for final assembly (Matuz 2002, p. 56).

Building a Shared Present and Future: Learnings from Henry Ford and. . .

19

Fig. 7 Packard Motor Car Company, building 10, Detroit Michigan. Exterior view. (Bucci 2002, p. 33) public domain

From the Wuity six-step process, Albert Kahn overcame [getting stuck] in using the traditional factory plant wood and masonry designs and was able to [let go] and test out new innovative techniques with steel-reinforced concrete for multistory buildings, roof window monitors for ventilation, and large sash windows for walls allowing more sunlight into the plant. The results were immediately visible with more open unobstructed floor space necessary for flexible manufacturing with large amounts of sunlight and ventilation. It took [mindful observation] of impact of these designs along with the [visual analogy] architectural drafting and drawings to realize these [insights] in the actual [implementation] and construction of these factory buildings in a very short time. From Gloor’s six indicators of high-performance teams, we can observe Albert Kahn’s [central leadership] skill of listening to his clients and rendering a factory plant design to meet their unique and growing factory plant needs. [Rotating leadership] was clear using the Kahn Bar-reinforced concrete, innovations of his brother Julius, which helped obtain a [balanced contribution,] with a [rapid response] through the speedy delivery of his final drawings and the actual construction of the factory plant. We need to assume an [honest language] exchange between Ford and Kahn based on a [shared context] of factory building requirements and the resources available to both men in the greater Detroit area.

20

K. Riopelle and X. Wang

Fig. 8 Packard Motor Car Company, building 10, Detroit Michigan. Interior view. (Bucci 2002, p. 33) public domain

While the Highland Park Plant was being built, Ford and his team continued to experiment with improving the manufacturing efficiency and cost reduction and conducted small assembly line production in subassemblies at the Piquette Plant beginning with the Model N. It took about 5 years of trial and error experimentation starting at the Piquette Plant in 1908 until 1913 when the “assembly line” was completely worked out, which was 3 years after the Highland Plant opened in 1910. Also, Ford’s team did not invent any of the processes that they finally used. They just combined all these processes in a unique way. They used their [mindfulness observation] and [visual analogy] skills to string it all together [gaining insights and implementation]. As Charles Sorensen stated in his book, My Forty Years with Ford, this was a time of considerable experimenting with mass production and the assembly line. Here he states his recollection of these times (Sorensen et al. 2006): . . . An equally slow evolution was the final assembly line, the last and most spectacular link in mass production. Both “just grew,” like Topsy. But whereas the car evolved from an idea, mass production evolved from a necessity; and it was long after it appeared that the idea and its principles were reduced to words. Today, we do not hear so much about “mass production” as we do about “automation.” Both evolve from the same principle: machine-produced interchangeable parts and orderly flow of those parts first to subassembly, then to final assembly. The chief difference is that mechanized assembly is more complete in automation; where men once tended machine

Building a Shared Present and Future: Learnings from Henry Ford and. . .

21

Fig. 9 Highland Park Plant Assembly building design (Bucci 2002, p. 38). Public domain tools, the job is now done electronically, with men, fewer of them, keeping watch over the electronics. Interchangeable parts were not new in 1913. Johann Gutenberg, the first printer in the Western world to use movable type, employed that principle 500 years ago. Eli Whitney used interchangeable parts when making rifles in the early days of the Republic; and in early days of this century Henry Leland, who later sold out to Ford, applied the same principle in the first Cadillac cars. Overhead conveyors were used in many industries, including our own. So was substitution of machine work for hand labor. Nor was orderly progress of the work anything particularly new; but it was new to us at Ford until, as I have already described, Walter Flanders showed us how to arrange our machine tools at the Mack Avenue and Piquette plants. What was worked out at Ford was the practice of moving the work from one worker to another until it became a complete unit, then arranging the flow of these units at the right time and the right place to a moving final assembly line from which came a finished product. Regardless of earlier uses of some of these principles, the direct line of succession of mass production and its intensification into automation stems directly from what we worked out at Ford Motor Company between 1908 and 1913. Henry Ford is generally regarded as the father of mass production. He was not. He was the sponsor of it. (pp. 115, 116.)

Henry Ford and his early team were a Collaborative Innovation Network. The team included James Couzens, “who controlled expenditures, organized sales, and set the pattern for business operations, . . . and drove the company into the new Highland Park Plant” (Sorensen et al. 2006, p. 36), Harold Wills in metallurgy and

22

K. Riopelle and X. Wang

Fig. 10 Highland Park Plant Complex aerial view 1908–1915 (Bucci 2002, p. 42). Public domain

tool design developing new steel alloys and their uses and setting up the Ford laboratory, Walter Flanders who was an expert at rearranging the machines on the shop floor, Charles Sorensen who was a patternmaker and Ford’s translator of his ideas into physical drawings and objects, and Ernest Liebold, Ford’s personal secretary, who wrote every letter and handled all of Ford’s correspondence. Together this diverse team moved Ford Motor Company from [getting stuck] in a highly skilled custom handmade automobile era to a new era of high-volume mass production and the continuous movement of material from one end of the plant to a finished product at the other end. The team [let go] of the idea of handmade custom, one-of-akind automobiles. The team members were [mindful observation] of moving work to the worker, rather than moving the worker to the work as a [visual analogy] from observation of practices in other industries. Visual clarity helped the team [gain insights] about how the substitution of unskilled labor using precision machines to make interchangeable parts could lead to a dramatic cost reduction. The team [implemented] these ideas through the sequencing of machines and the flow of subassemblies into a final product, a car. Henry Ford was the [central leader] but also engaged in [rotating leadership] with each of the COIN members making a [balanced contribution] within their areas of production, sales, and cost reduction. Their responses were [rapid] because the team members were in the factory, on the shop floor, and could see and measure the results of their efforts and experiments. Their language was [honest,] even if harsh and loud

Building a Shared Present and Future: Learnings from Henry Ford and. . .

Fig. 11 Highland Park Plant exterior view (Bucci 2002, p. 43). Public domain

Fig. 12 Highland Park Plant interior view (Bucci 2002, p. 45). Public domain

23

24

K. Riopelle and X. Wang

at times, but they shared the same [context] and wanted the high-volume, low-cost production to work. In 1914, after 4 years of experiments with mass production and the assembly line, the Highland Park Plant was in full swing, and Ford made 248,307 Model T’s (Matuz 2002, p. 57). With sales skyrocketing the price was dropping from $780 down to $360. Stockholders profits rose from 4.1 million in 1910 to $57.1 million in 1916 (PBS American Experience Features Henry Ford). But Henry Ford made another groundbreaking business decision on January 5, 1914, and announced the $5 a day wage for an 8-h work day, which was more than twice the current pay rate of $2.34 a day. The reaction was swift and visible with 10,000 job seekers showing up at the Highland Park Plant the next day (PBS American Experience Features Henry Ford). Ford’s vison was for the common man to afford to buy his car. The problem was the common man did not make enough money to buy his car. However, Ford made the [mindful observation] that as his sales rose and prices came down, he could raise wages and still make a profit. He had the [insight] that if he could raise the workers’ daily wage, the immediate implication was that the common man could in turn afford to buy his Model T. Charles Sorensen works out the sales, profits, and wage forecast on a blackboard to illustrate this was a real case. In summary, Ford followed the Wuity six-step process. He was [stuck,] how could a common man afford his car? He then [let go] of conventional business wisdom of paying workers a low wage to keep costs down and profits up. He applied the [mindful observation] that as sales rose, prices could drop, but he could still make a profit on higher volumes and raise wages, and in turn the worker could realize that he could buy Ford’s Model T. Given this insight he [implemented] the idea of raising a worker’s wage to $5 a day. Nevertheless, success brought back the reoccurring problem; they needed a bigger plant. In 1915, Ford and Kahn were back designing and building another new plant named the Rouge Plant, which is our next factory case.

4.2

The Rouge Plant

In 1912–1913, just 2 years after the opening of the Highland Park Plant, Henry Ford had Charles Sorensen make production estimates for the hot selling Model T. The projected volumes made it clear that by the year 1920, their new plant was going to be too small to produce an estimated 1 million vehicles (Sorensen et al. 2006, p. 137). Thus, Henry Ford started to look for land for an even bigger and more ambitious plant. The Highland Park Plant had surfaced some new and unexpected land problems. First, there was the need for space for employee parking; second, there was a need for more space for inventory waiting for shipment; and third, there was more space needed for the plant to smelt its own iron ore and making its own steel. Ford Motor Company needed land, flat land, and access to the Great Lakes for importing iron ore, limestone, and other raw materials to make the steel that was a

Building a Shared Present and Future: Learnings from Henry Ford and. . .

25

major cost component of the Model T automobile. Therefore, Ford ultimately purchased 2000 acres of land along the Rouge River, where it emptied into the Detroit River and traveled up to his home town of Dearborn, Michigan (Sorensen et al. 2006, p. 154). Ford’s vision was clear: own all the raw materials; do not be dependent upon unreliable suppliers who may disrupt production; eliminate all the middle men; and keep lowering the costs. Ford had one more mindful observation—the multistory factory building was inefficient. There was too much waste in moving parts and subassemblies up and down and in between floors. Ford wanted a new single-story plant under one roof, and Albert Kahn was called upon again to design and build this new Rouge Plant. Ford: “Albert, . . . I need you to design a new factory. This time, I want all production, from start to finish, to take place not just under the same roof, but on the same floor” (Matuz 2002, p. 59). He explained that he wanted the plant to be entirely self-sufficient, from making steel and glass to assembling cars. Albert Kahn was prepared for this request, drawing upon his factory design experience from four previous contracts. First, in 1904 Kahn had designed and constructed a one-story sawtooth-roofed plant for the Burroughs Adding Machine Company in Detroit. Second, in 1906 Kahn had helped to co-design the Pierce Great Arrow Automobile Plant in Buffalo, New York, which was a set of seven one-story buildings foreshadowing Ford’s assembly line production. See Fig. 13 for the building plan for the Pierce Great Arrow Automobile Plant in Buffalo, New York (Bucci 2002).

Fig. 13 Pierce Great Arrow Automobile Plant in Buffalo, New York, building plan. (Hildebrand 1974, p. 37) Public domain

26

K. Riopelle and X. Wang The layout was articulated in seven one-story buildings of reinforced concrete with sheds [sic] roofs, each corresponding to one segment of the production process: Garage, Manufacturing Building, Assembly Building, Body Building, Brazing Building, Motor Testing Building, and Power House. The raw materials arrived via railroad to the Brazing Building; from there they proceeded towards the Manufacturing Building (that produces motors and frames) and the Body Building. The Assembling Building was located in the center of the of the plant where the three parts of the car were assembled (The motors first passed through the Motor Testing Building for inspection); once completed, the vehicles were sent to the Garage to be equipped. This model of organization of the production flow, one that moved the piece work through single buildings, anticipated the concept of the assembly-line that would be more extensively and systematically employed by Ford. . . (p. 35)

Third, in 1911 Kahn had used steel for the sawtooth-roof machine shop for Ford at the Highland Park Plant and, fourth, for the Packard Forge Shop. Thus, Kahn drew upon these four previous building designs to begin his work for designing the Rouge Plant Complex of buildings—all one story, steel framed, with sawtooth roofs for ventilation and glass walls enabling sunlight to light the interior. Nearly all of Kahn’s works after 1915 for Ford were steel structures (Hildebrand 1974, p. 92). There were three significant advantages to building a single-story building: lower costs, speed of construction, and flexibility of use of the interior space. A single-story building was cheaper to build because it did not require the strength of steel-reinforced concrete columns to support the weight of multiple floors; the time-consuming construction method of forming the columns, pouring the cement, waiting for it to be cured, and then removing the forms could be eliminated, which lowered the costs and sped construction. Last, a single-story-building provided more flexibility of reworking the interior space for expansion. Moreover, Ford had the [mindful observation] . . . (Hildebrand 1974). that smooth rapid internal operations meant little unless the materials could be brought to the plant and the product taken away with the same ease. For the Rouge site was ideal; it served [sic] by major rail lines, including a direct connection to Highland Park, and also by the river itself, which linked Rouge to the vital Great Lakes ports. (p. 93)

World War I (July 28, 1914–November 11, 1918) was raging at this time, and Ford convinced the US government that he could build the Eagle Submarine Chaser Plant on the Rouge site and mass produce the Eagle Boat just like a Model T. On January 17, 1918, he was given the approval to proceed with building the boat. Five months later in May of 1918, the plant was in operation and launched the first boat on July 10. The factory plant construction was four times faster than the Highland Park Plant, thanks to the savings using steel in a single-story factory (Hildebrand 1974, p. 93). “The drawings were dated February 11, 1918, less than four weeks after Ford was directed to proceed. The plant was complete in fourteen weeks of the issuance of the drawings.” (Hildebrand 1974, p. 99) The Eagle Plant consisted of five immense aisles, each 51 feet wide and 1700 feet long, all under a single roof. This was the largest single area Kahn had designed to date. If you multiple the five aisles by 51, it equals a total interior of 251 feet as compared to the 85-foot width for the Highland Park Plant, an increase in width of 166 feet or 195%. The Eagle Plant length of 1700 feet was 840 feet or 98% longer

Building a Shared Present and Future: Learnings from Henry Ford and. . .

27

Fig. 14 Rouge Plant Section and Elevation drawing (Bucci 2002, p. 49). Public domain

than the 860-foot length for the Highland Park Plant. Two of the five aisles were dedicated to railroad tracks for the movement of parts and materials. The other three aisles were for the boats and their sequence of building. The building was 30 feet 8 inches tall except for the last 400 feet when the height grew to 50 feet 9 inches of three boat aisles where the finished boat needed the extra height to exit the plant floor. This gave the exterior look of the building to be a three- to five-story building, but in fact it was a single story. When the boats exited the plant, they were placed on a transfer table to a launching slip 600 feet away directly into the Rouge River with access to the Great Lakes (Hildebrand 1974, p. 93). Figure 14 is a section and elevations drawing of the Rouge Plant. Figure 15 is an aerial view of the Rouge Plant. Figure 16 is an interior view. Figure 17 is a picture of the last Eagle Submarine Chaser Boat exiting the plant in 1919. The factory was then transformed again for a return to automobile production (Matuz 2002, p. 58). Figure 18 is a picture of an Eagle at sea. The next and final case will be the Willow Run Bomber Plant designed to fulfill the urgent need in the United States for airplane bombers to win World War II.

4.3

The Willow Run Bomber Plant

On December 7, 1941, Japanese war planes bombed Pearl Harbor in the Hawaiian Islands and nearly destroyed the entire US Pacific Naval Fleet. The United States

28

K. Riopelle and X. Wang

Fig. 15 Rouge Plant aerial view (Hildebrand 1974, p. 97). Public domain

Fig. 16 Rogue Plant interior view (Hildebrand 1974, p. 98). Public domain

Building a Shared Present and Future: Learnings from Henry Ford and. . .

29

Fig. 17 Rouge Plant with completed Eagle Boat Exiting Plant (http://www.navsource.org/) public domain http://www.navsource.org/archives/12/12040106.jpg

Fig. 18 An Eagle Submarine Chaser at Sea (http://www.navsource.org/) public domain http:// www.navsource.org/archives/12/12041906.jpg

suddenly entered World War II and found itself in short supply of military bomber planes that it needed immediately. Once again, the Ford Motor Company was called upon to help. William S. Knudsen, Sorensen’s former deputy at Ford who became president of General Motors, was heading up the production side of rearmament for the US Defense Department at the time and wanted to know if Ford could help build a bomber engine, speeding up production of the Air Force’s B-24 bomber.

30

K. Riopelle and X. Wang

On January 8, 1941, Charles Sorensen flew out to Consolidated Aircraft in San Diego, California, and toured the bomber factory whose goal was to produce one bomber a day or 350 a year. At this slow pace, it would take 3 years to produce the 1000 four-engine bombers that the United States needed. What he saw was a custombuilt plane without using any of the principles of mass production. The final assembly occurred outdoors, which caused temperature differences to distort the pieces and keep them from fitting together, and that in turn caused further delay. Sorensen went to his hotel room that night and sketched out the mile-long Willow Run Bomber Plant, including a cost estimate for producing a B-24 at a rate of one an hour rather than one a day (Sorensen et al. 2006). I knew the same fundamentals applied to high-volume production of both, the same as they would to an electric egg beater or to a wrist watch. First, break the plane’s design into essential units and make a separate production layout for each unit. Next, build as many units as are required, then deliver each unit in its proper sequence to the assembly line to make one whole unit—a finished plane. To house all this and provide for efficient operation there should be a new plant specially designed to accommodate the progressive layout. I saw no impossibility in such an idea even though mass production of anything approaching the size and complexity of a B-24 never had been attempted before. (p. 281)

He explained: I was back at my old game of sketching a series of manufacturing and subassembly operations and their orderly progression toward becoming major units—a game I had played many times since that morning in 1908 at the Piquette Avenue plant when we first experimented with a moving assembly line. Again, I was practicing my production planning philosophy, which stemmed from my patternmaking days when I fashioned wooden models of Henry Ford’s half-thought-out designs: “Unless you see a thing, you cannot simplify it. And unless you can simplify it, it’s a good sign you can’t make it.” As I look back now upon that night, this was the biggest challenge of my production career—bigger than any Model T assembly line sequence for Highland Park, more momentous than the layout and construction of the great River Rouge plant in which I’d had a part. It took 8 years to develop Ford mass production system, and 8 more years before we worked up to a production of 10,000 cars a day. Now, in one night, I was applying 35 years of production experience to planning the layout for building not only something I had never put together before, but the largest and most complicated of all air transport and in numbers and at a rate never before thought possible. Once again I was going on the principle I had enunciated many times at Ford: “The only thing we can’t make is something we can’t think about.” (pp. 281–282) . . .Standing over the papers, I roughed out on Coronado Hotel note paper a pencil sketch of the floor plan of a bomber plant. It would be a mile long and a quarter mile wide, the biggest single industrial building ever. (p. 283)

The next morning Sorensen proposed to produce the whole bomber in a new plant at a rate of one an hour and presented his detailed plant sketched and at an estimated cost of $200 million. The Air Force accepted. Of course, the next call was to Albert Kahn to design and build the plant based on Sorensen’s plant floor sketch. Just like the Rotunda Plant, raw materials were delivered to one end of the plant, and the finished planes emerged out the other end all under one roof. Once the planes were completed, they were flown off on an adjacent runway away from the plant. However, the Rotunda Plant was conceived and built over years; now the timetable was shrunk to months and days.

Building a Shared Present and Future: Learnings from Henry Ford and. . .

31

This was their timetable (Sorensen et al. 2006). February 25, 1941, project approved in Washington. April 18, 1941, official ground-breaking ceremony. May 3, 1941, first structural steel erected. August 12, 1941, first machine tools installed. November 15, 1941, first production man-hours. September 1942, first acceptance of B-24s completely assembled at Willow Run. From approval of project to building the world’s largest manufacturing structure and delivery of planes—nineteen months! (Sorensen et al. 2006, p. 286)

There was one important difference in this factory plan design—no windows because of the need for wartime blackouts when lights had to be extinguished. Therefore, Albert used artificial light throughout the plant. Albert Kahn stated that “speed” was the key characteristic for National Defense (Bucci 2002). . . . speed is the password of the Defense Program. The decisions to construct or enlarge must be made quickly, and the architectural designs must be ready immediately. . .. There is no time for philosophy, to wait for inspiration or to consider the aesthetic problems. . .. Simplicity of design and construction is imperative. Every day counts and minutes must be saved. It’s not only a question of dollars and cents, but, today it is a question of life and death. (p. 106)

In summary, Sorensen’s six-step Wuity process was very familiar even though the task of manufacturing a bomber airplane was an unfamiliar task. The contractor was [getting stuck,] in a slow custom build process. Sorensen [let go] of the idea that people usually think cars and planes are completely different and needed to follow a different production process. He had the [mindful observation] and [visual analogy] that this production process was just like building a Model T or an Eagle Submarine Chaser Boat; he just needed to plan a similar layout of the build process. With this anti-intuitive [insight] that the same fundamentals applied to high-volume production of a car, a boat, and an airplane, he sketched out the entire Willow Run Bomber Plant, costed it, and said the Ford Motor Company was ready to [implement] and build a plant to produce a bomber an hour rather than one a day. From Gloor’s six indicators of high-performing teams, Sorensen was a clear [central leader,] with the assumption of [rotating leadership] and [balanced contribution] between Ford and Kahn on how to build the Willow Run Bomber Plant. Sorensen’s [“rapid response”] was within a single night to design the plant and cost it. He used [honest language] that Ford Motor Company could do it and they had a [shared context] of urgency to get these planes built quickly. Figures 19, 20, and 21 depict three views of the Willow Run Bomber Plant. Figure 19 shows the layout of the plant floor. Figure 20 is an aerial view including the airport runways, and Fig. 21 is an interior view showing the planes coming off the assembly line. Both Ford and Kahn used their Co-Wuity and high-performance team experiences to learn from each other over a span of 34 years and some 1000 buildings. They each [got stuck,] and [let go] of traditional thinking, using their [mindful observation] of the situation, along with a [visual analogy] to [gain insights] into [implementing] a novel solution. When they got stuck again, they just repeated this

32

K. Riopelle and X. Wang

Fig. 19 Willow Run Bomber Plant floor diagram (For more pictures of the Willow Run Bomber Plant, see https://www.google.com/search?q¼Pic:Willow+Run+Bomber+plant&rlz¼1C1CHBF_ enUS739US739&tbm¼isch&source¼iu&ictx¼1&fir¼HR2wdvArG_hHuM%253A% 252CX3D9YKqjqPWhRM%252C_&usg¼__CjBvHAgtw7C5JwnIXNAA3yC2MuQ%3D& sa¼X&ved¼0ahUKEwj60aGJsIrbAhVi04MKHXTKBxsQ9QEINjAB#imgrc¼ptifiTJVbkBkRM:) public domain

Wuity six-step process to find another solution. The three historic case studies of the design and building of the Highland Park Plant for the Model T, which produced an affordable car for the common man, the Rogue Plant to build the Eagle Submarine Chaser Boat for World War I, and, last, the Willow Run Bomber Plant used to make B-24 Bombers for World War II at a rate of one an hour, are testaments to the strength and versatility of this Co-Wuity and high-performance team relationship. Thus far we have reviewed how Albert Kahn helped Henry Ford achieve the goal of mass production and the assembly line process by designing factory buildings which could accomplish that task. The next section describes the reciprocal relationship of how Henry Ford helped Albert Kahn achieve his goal of a high-volume architectural firm for custom building plans.

5 Mindful Social Interaction and Shared Present and Future Albert Kahn’s diverse client relationships with the automobile companies, and especially Henry Ford, provided him with an upfront, close, and insightful perspective about how mass production was being conceived, organized, structured, and

Building a Shared Present and Future: Learnings from Henry Ford and. . .

33

Fig. 20 Willow Run Bomber Plant aerial view (https://www.thisdayinaviation.com/tag/willowrun-bomber-plant/) Public Domain https://www.thisdayinaviation.com/tag/willow-run-bomberplant/

built. Starting with his first automotive client James Joy and the Packard Motor Car Company, he learned about the importance of building design for the flow of material through a factory building. This was reinforced with the joint architectural work on the Pierce Great Arrow Automobile Plant in Buffalo, New York, in 1906 that illustrated the segmenting and sequencing of work across seven buildings. Henry Ford’s contract to build the Highland Park Plant enabled Kahn to listen to Ford’s vision of how he planned to create an assembly line for a high-volume and low-cost automobile. The Model T and the Rouge Plant contract in 1918 reinforced how Kahn was to organize and structure his own office using many of the principles he had been exposed to. The automobile companies were designing and building a physical product—an automobile. Khan was designing and delivering a knowledge product—a building plan to be produced. Kahn’s office staff grew from just a few people in 1901, to 40 in 1910, to 80 in 1918, to 200 in 1923, and by 1935 the staff had grown to between 400 and 600 (Hildebrand 1974, pp. 6, 126, 153). Using the Wuity six-step process, Kahn found himself [getting stuck,] asking himself how do I complete so much work? He [let go] of the conventional architectural firm practice of just delivering an artistic building set of drawings, but instead had the [mindful observation] to assemble a diverse, multiskilled team to collaborate

34

K. Riopelle and X. Wang

Fig. 21 Willow Run Bomber Plant interior view https://nl.wikipedia.org/wiki/Willow_Run#/ media/File:B-24_bomber_at_Willow_Run.jpg (public domain)

using the [visual analogy] of a mass production assembly line to organize the complete design planning process or sequence of steps from start to finish. The collaborative team gained [insights] about how to deliver such a comprehensive and harmonious plan for its [implementation] or construction for high volume and speed (Bucci 2002). Albert Kahn stated in an article “Architectural Trend,” published in 1931, declared the end of the era “of individualist, the temperamentally[sic] artist,” [letting go] replacing this figure with “the collective efforts of group of men cooperating [rotating leadership] under proper direction.” [central leadership] (pp. 126–127)

Figure 22 represents Albert Kahn Inc.’s 1918 office floor plan in the Marquette Building, Downtown Detroit, and, in Fig. 23, the company’s organizational structure of the office team. The Kahn office layout [visual analogy] was organized like Ford’s Highland Park Plant for high volume and [rapid response] as described in a column by G.C. Baldwin in The Architectural Forum, vol. 29, no 5, November 1918, (pp. 125–126). As follows (Ibid): In addition to executive offices, an atrium, corridors, underground passageways, facilities, sample rooms, dressing rooms, meeting rooms, and a library, there are two large design rooms. There are also separate rooms for the mechanical and structural engineers, one area

Building a Shared Present and Future: Learnings from Henry Ford and. . .

35

Fig. 22 Albert Kahn’s office structure floor plan in the Marquette building 1918 (Bucci 2002, p. 126). Public domain

Fig. 23 Albert Kahn’s organizational chart. https://upload.wikimedia.org/wikipedia/commons/ thumb/f/f6/Kahn_firm_layout_1939.jpg/1000px-Kahn_firm_layout_1939.jpg for design, two areas for specific technicians, one for the compilers of these specifics, a separate room for the typists, offices for the head superintendent and for the field superintendent, a room for estimators, and two places for filing the contracts and correspondence. The offices for the executives and the meeting rooms are arranged along one side of the building, the design rooms along the other. The mechanical engineering and the structural

36

K. Riopelle and X. Wang engineering departments are situated on opposite corners. . . . The superintendents’ offices are arranged in two groups on both sides of the meeting room. The hallways that serve them both are floored with artificial, sound proof tiles . . . The designers’ rooms are separated by clear glass divisions. The offices are enclosed walls, and every office is equipped with a telephone for the city, an internal telephone and an intercom. (p. 128)

This office layout has been designed for maximum efficiency and clusters people close to one another for ease of communication, [rotating leadership, balanced contribution] including equipping everyone with three modes of electronic communication [rapid response]. Kahn had devised a method for determining the status of any given project [central leadership] as reported by G.C. Baldwin as follows (Hildebrand 1974): All departments [balanced contribution] have graphical progress reports on their work [shared context]. At the beginning of any work the estimated progress was indicated by a curve in black ink on a co-ordinate paper, and the actual progress recorded from day to day in red ink [honest language]. Any marked divergence in these curves indicates serious delay, and daily inspection of the records enables prompt action to be taken to remove it [rapid response]. The superintendent’s reports [sic] the progress in this way and also by means of daily and weekly reports [balanced contribution]. Thus, a general and detailed supervision of all work is possible from the executive offices [central leadership]. (p. 60)

Hildebrand cites how this office work process parallels Ford Motor Company’s Highland Park operations (Ibid): Eighteen different paperwork forms [shared context] were used for job records, not including general records and interoffice communications [honest language]. These forms were similar to those used by the Ford Motor Company in its operations at Highland Park. The entire organization of the work progress . . . recalls the automobile industry’s ‘complex system for charting stock inventory, the transfer of materials, job routing, and the precise state of each assembly-line [balanced contribution and honest language] at a given moment [central leadership].’ (p. 60) Kahn has been described as a “rigorous administrator” and an “arduous taskmaster” who demanded an intense team effort from all his employees. He did not tolerate “backbiting” and if someone complained about someone else he and the defendant were taken to Kahn’s office for the dispute to be settled and dropped [honest language]. “Prima donnas repressed their egos or were dismissed” Kahn saw himself as a conductor of a symphony [central leader] and the objective of the talents [rotating leadership] were toward an end impossible to achieve by and single participant. (pp. 126–127)

From the point of view of Gloor’s six indicators of high-performing teams, Kahn’s office exhibits a high match. Kahn was a strong [central leader,] who demanded the people contribute their expertise ensuring a [rotating leadership] and [balanced contribution] among the multidisciplinary team members. [Rapid response] was the expectation to close any gaps in the projects’ performance time using face-to-face communication in team meetings or one of the three electronic means of communication. It is assumed the team members used [honest language] to resolve conflicts because they had a [shared context] to deliver a building that worked for the client’s needs. This office system was put to test in 1929 when Kahn was contracted to rebuild Russian factories in 4 years between 1929 and 1933. On May 7, 1929, a New York

Building a Shared Present and Future: Learnings from Henry Ford and. . .

37

Times Headline read: “America to Build Soviet Auto Plants.” Albert Kahn Associates sets up offices in Moscow, headed by his brother, Moritz Kahn. Moritz supervised the design of more than 500 buildings including steel, automotive, tractor factories, as well as airplane and chemical plants with the support of the home office in Detroit. To design so many buildings so fast, over 1500 Soviet draftsmen were employed. They worked during the day and attended classes at night to learn the techniques of Albert Kahn (Matuz 2002, p. 81). In this way, Albert Kahn set up a Collaborative Learning Network for the Russian government to help them quickly rebuild. At the same time, Charles Sorensen from Ford Motor Company negotiated with the Russians some 40 million dollars in business from 1929 to 1936 to teach the Russians how to build auto and steel plants based on the Rouge Plant (Sorensen et al. 2006, p. 193). In summary, Henry Ford and Albert Kahn are an example of a Co-Wuity relationship whose high-performance teams extended their Collaborative Innovation Network (COIN) to a Collaborative Learning Network in helping to train Russians and rebuild Russia after World War I. The next section is a reflection on what these three case studies can provide as lessons learned for today’s emerging technologies and entrepreneurs.

6 Lessons Learned for Today’s Entrepreneurs In building a shared future, what can we learn from two start-ups and their founders, Henry Ford of the Ford Motor Company and Albert Kahn of Albert Kahn Architects, back in 1896 some 122 years ago? What are the key aspects of their personalities and Co-Wuity relationship of 34 years along with their high-performance teams that today’s entrepreneurs and their start-ups can learn from and benefit from today? Both founders, Henry Ford and Albert Kahn, had developed the very keen personal quality of mindfulness and applied this skill over and over in their respective business areas and shared the results of their talent with one another. Henry Ford was relentless in his pursuit of a high-volume low-cost car for the common man. This led him and his team to be mindful and use a wide range of existing practices in a unique combination to form a whole process, what is called today mass production. The moving assembly line that moved work to the workers rather than workers moving to the work, sequencing machines that made standardized parts, purchasing in large quantities to smooth production, and avoiding slowdowns or stoppages due to part shortages, vertically integrating all the raw materials of iron ore, steel, glass, rubber, and other parts, all designed to produce the seamless, continuous movement of material. Ford was mindful that the new transportation networks of rail, waterways, and roads were key to the continuous movement of material in, through, and out of his plants. All these components needed to work seamlessly together as a whole. This was a time when Ford and his team could visibly see everything on the plant floor,

38

K. Riopelle and X. Wang

whether the production was working or not; there was simple measurement of production output, at what cost, and profit. Albert Kahn likewise honed his mindfulness skill in the observation of the relationship between a building’s interior space and exterior construction. His life was devoted from early childhood through adulthood to observing by drawing and sketching a wide variety of buildings and architectural styles. But it was not just his talent for drawing fast and with great detail alone that was so important. He had developed an intuitive understanding of how interior space and its construction worked. Kahn had studied a wide variety of architectural styles developed over the centuries beginning with his early European scholarship tour and later through his personal trips to Europe and his constant contact with other artists and architects and also through his professional and personal social networks and relationships, including with his wife, all of whom kept him in a constant learning mode. Thus, the first lesson learned for today’s entrepreneurs is to develop their mindfulness skill and apply it to their start-ups. Both Ford and Kahn had a well-developed mindfulness skill, and they applied it over and over again throughout their lives. Second, both Ford and Kahn were visual thinkers and used visual analogies to innovate and experiment in their businesses. Ford could not read a blueprint, but found a talented young man, Charles Sorensen, who was a patternmaker to translate his ideas into physical objects so he could better understand the attributes and qualities of an object or idea. Kahn was color blind, but this did not prevent him from seeing the essence of a building and the relationship among the interior and exterior spaces and how they “felt” and worked together. Ford and his team found practices in other industries, such as meat packing and mill houses, to use as visual analogies and adopted similar practices in his automobile manufacturing. Kahn could see how a client’s building desire could be fulfilled by using a visual analogy to form a design and a building plan. Both men used their ability to visualize situations using analogy to gain insights and implement those ideas as concrete plans. Thus, each used the Wuity six-step process again and again throughout their careers. Therefore, the second learning for today’s entrepreneurs and their start-ups is to learn and practice visual analogy. The third lesson learned is Wuity + Wuity is a powerful combination. Henry Ford and Albert Kahn collaboration demonstrated groundbreaking results in automobile manufacturing and architectural factory design. Kahn helped Ford solve the issues in factory building design of obstructed space, poor ventilation, limited light, and fire hazard. Ford helped Kahn solve the issue of high-volume throughput of work in a sequential standardized but flexible work process. Ford’s Wuity focused on a physical product or deliverable, while Kahn’s Wuity focused was on a knowledge product of building design plan. The fourth lesson learned is that each founder, Ford and Kahn, relied on a talented team of individuals who could work together within their start-ups and with others outside their companies. They did not succeed on their own talent and skills but needed the collaboration of a diverse set of individuals working together toward a shared goal. Ford’s team included James Couzens, Harold Wills, Walter Flanders, Charles Sorensen, and Ernest Liebold. Kahn’s team included his three brothers,

Building a Shared Present and Future: Learnings from Henry Ford and. . .

39

Julius, Moritz, and Louis, along with architects Ernest Wilby and Wirt Rowland to name just a few. The fifth lesson learned is that Gloor’s six high-performance indicators neatly fit the attributes of Ford and Kahn’s historical cases some 122 years ago, and we argue they remain relevant for today’s start-ups. We suggest that entrepreneurs and their start-up teams ask themselves six questions: (a) (b) (c) (d) (e) (f)

Does our team have strong central leadership? Do we have evidence of rotating leadership? Do we have balanced contribution among our team members? How fast is our response to one another, our clients, and our suppliers? Do our team members use honest language in our communication? Have we developed a shared context?

The sixth lesson learned is that the conditional context of people’s lives and their times, including the particular business conditions and families and friends, can often have a significant impact on the entrepreneur and push start-ups in one direction or another. The nineteenth-century State of Michigan’s historic growth and development in its transportation infrastructures of railroads, waterways, roads, and new industries of steel, railroad car construction, mining and shipping, and the communication media of newspapers and telephone, the expansion of education, and the founding of institutions of higher education, along with rapid immigration and diverse population increase, all combined to provide the rich business conditions for Ford and Kahn to build their businesses and not only grow but thrive. For Ford, the availability of large tracts of flat land was especially important as the locations for his expanding factory needs. For Kahn, the need for new architectural building plans was in high demand for not only factories but homes, offices, cultural sites, and other buildings. Two World Wars brought Ford’s and Kahn’s business expertise and teams to the forefront for urgent and immediate attention to pressing problems. We ask today’s entrepreneurs and their start-up teams to look around at their conditional contexts and business conditions and ask what resources are growing and can provide advantage and leverage in their endeavors. Would an unexpected crisis thrust their talents, expertise, and teams into the forefront? In addition, the conditional context for entrepreneurs includes their family and friends who may have an important influence on their development. Ford had his friend Charles B. King who gave him parts to build his first quadricycle, and Thomas Edison encouraged him to keep pursuing his horseless carriage dream. For Kahn, his mother and father were always encouraging him to draw and sketch; buying him pencils, paper, and a drafting board; and networking him to family friends who were artists, sculptures, and architects for lessons and jobs. Last, but certainly not least, both Henry Ford’s wife, Clara, and Albert Kahn’s wife, Ernestine, were prominent confidents to whom they listened. Both men took their wives’ advice seriously on many issues and situations. We ask today’s entrepreneurs to reflect and assess their family and friends’ networks to recognize and identify those who support them and encourage them as well as serve as confidants during difficult or tumultuous times or situations.

40

K. Riopelle and X. Wang

7 Conclusion and Future Research In conclusion, this paper’s contribution to the literature is an interweaving of two streams of research based on Gloor’s six directional indicators of high-performance teams and the six process steps of Wang’s Wuity using an historical archival case analysis method as exemplified by Weick and Sutcliffe. We believe that this is a fruitful line of scholarly inquiry to use the reexamination of historical events to elicit guidelines and principles for today’s entrepreneurs and start-ups for building a shared present and future. Wuity is a higher cognition with the flexible and imaginative ability to gain intuitive understanding of underlying phenomena and connect deliberative thinking for insightful problem-solving through a process of mindful observation and visual analogy. Mindfulness has a power to enable people to develop a thinking model for creativity and innovation for societal impact. Wuity makes a use of imagery, metaphors, and analogy rather than concepts, logics, and analysis. Given Wuity as a style of thought and information processing, we find in high-performance teams Co-Wuity interaction manifests as entangled minds through COINs memberships. This entanglement is a spontaneous mind-to-mind and mind-to-matter interaction. The entanglement of minds or their mindful social interaction has three distinct features. The first feature, from a mindfulness perspective, is that team leaders and members don’t invent anything, but they do innovate the next big thing. The second feature of their mindful social interaction is taking on another’s perspective. Looking from the point of view of another is altruistic in a broad sense. Participants are using their Wuity visual thinking to build a bigger picture for team members to co-create. They are doing their parts for themselves, for others, and for the community. The third feature is through mindful inquiry, mindful observation, mindful learning, and visual thinking, leaders and their teams enable to entangle their minds for co-creating a higher performance for a shared present and future. Our contribution is expanding Wuity research from the intrapersonal or individual level, to the interpersonal level and to the larger collective level, which is the social interaction through networks. Finally, we recommend the continued development of more case studies for theory building connecting COINs, Wuity, and mindfulness research. Specifically, we recommend exploring cases where entrepreneurs have faltered and become stubborn and rigid as was the case for Henry Ford when he famously stated: “Give them any color they want so long as it is black” (Sorensen et al. 2006, p. 217). Exploring example cases that illustrate both how Wuity thinking produces transcendental insights and benefits to its practitioners and how suspension of mindfulness and collaboration can produce narrow tunnel thinking and negative effects will ultimately strengthen theory development about COINs.

References Bucci, F. (2002). Albert Kahn: Architect of Ford. New York, NY: Princeton Architectural Press. Curcio, V. (2013). Henry Ford. Oxford: Oxford University Press.

Building a Shared Present and Future: Learnings from Henry Ford and. . .

41

Dane, E. (2011). Paying attention to mindfulness and its effects on task performance in the workplace. Journal of Management, 37(4), 997–1018. Detroit Historical Society Timeline. (1787–1820, 1820–1860, 1860–1900). Accessed November 12, 2018, from https://detroithistorical.org/learn/timeline-detroit/early-american-detroit-17871820, https://detroithistorical.org/learn/timeline-detroit/boomtown-detroit-1820-1860, https:// detroithistorical.org/learn/timeline-detroit/industrial-detroit-1860-1900 Gloor, P. A. (2017a). Sociometrics and human relationships: Analyzing social networks to manage brands, predict trends, and improve organizational performance. Bingley: Emerald Publishing. Gloor, P. A. (2017b). Swarm leadership and the collective mind: Using collaborative innovation networks to build a better business. Bingley: Emerald Publishing. Gluesing, J. C., Alcordo, T., Baba, M., Britt, D., Harris Wagner, K., McKether, W., et al. (2003). The development of global virtual teams. In Virtual teams that work: Creating conditions for virtual team effectiveness. San Francisco: Jossey-Bass. Goleman, D., & Davidson, R. J. (2017). Altered traits: Science reveals how meditation changes your mind, brain, and body. Glendale, CA: Avery. Hildebrand, G. (1974). Designing for industry: The architecture of Albert Kahn. Cambridge, MA: The MIT Press. Hodges, M. H. (2018). Building the modern world: Albert Kahn in Detroit. Wayne State University Press. http://www.navsource.org/. http://www.navsource.org/ Industrial Detroit. (1860–1900). Accessed May 8, 2018, from https://detroithistorical.org/learn/ timeline-detroit/industrial-detroit-1860-1900 Langer, E. J. (1989). Mindfulness. Reading, MA: Addison-Wesley/Addison Wesley Longman. Matuz, R. (2002). Albert Kahn: Builder of Detroit. Detroit, MI: Wayne State University Press. Nevins, A., & Hill, F. E. (1954). Ford: The times, the man, the company (Vol. 1). New York: Charles Scribner’s Sons. PBS American Experience Features Henry Ford. http://www.pbs.org/wgbh/americanexperience/ features/henryford/; http://hfha.org/ford-timeline/ Smith, M. G. (2017). Designing Detroit: Wirt Rowland and the rise of modern American architecture. Detroit, MI: Wayne State University Press. Sorensen, C. E., Williams, S. T., & Lewis, D. L. (2006). My forty years with ford. Detroit, MI: Wayne State University Press. Wang, X., & Gloor, P. A. (2018). Wuity as higher cognition combining intuitive and deliberate judgments for creativity: Analyzing Elon Musk’s way to innovate. In F. Grippa, J. Leitão, J. Gluesing, K. Riopelle, & P. Gloor (Eds.), Collaborative innovation networks: Building adaptive and resilient organizations (pp. 165–182). Cham: Springer. Wang, X., & Li, P. P. (2017). Conceptual framework of “Wu”-based innovation at the team level: A case study on indigenous innovation in the context of aerospace projects. Foreign Economics and Management, 39(3), 3–23. Wang, X., & Li, P. P. (2018). Wu-based thinking integrating the east with the west: The analysis of Elon Musk’s four metaphors. Foreign Economics and Management, 40(1), 124–140. Weick, K. E., & Sutcliffe, K. M. (2001). Managing the unexpected: Assuring high performance in an age of complexity. Hoboken: Wiley. Weick, K. E., & Sutcliffe, K. M. (2011). Managing the unexpected: Resilient performance in an age of uncertainty. Hoboken: Wiley. Weick, K. E., & Sutcliffe, K. M. (2015). Managing the unexpected: Sustained performance in a complex world. Hoboken: Wiley.

Mirror, Mirror on the Wall, Who Is Leaving of Them All: Predictions for Employee Turnover with Gated Recurrent Neural Networks Joao Marcos de Oliveira, Matthäus P. Zylka, Peter A. Gloor, and Tushar Joshi

Abstract Employee turnover is a serious issue for organizations and disrupts the organizational behavior in several ways. Hence, predicting employee turnover might help organizations to react to these mostly negative events with, e.g., improved employee retention strategies. Current studies use a “standard analysis approach” (Steel, Academy of Management Review 27:346–360, 2002) to predict employee turnover; accuracy in predicting turnover by this approach is only low to moderate. To address this shortcoming, we conduct a deep learning experiment to predict employee turnover. Based on a unique dataset containing 12 months of time series of e-mail communication from 3952 managers, our model reached an accuracy of 80.0%, a precision of 74.5%, a recall of 84.4%, and a Matthews correlation coefficient value of 61.5%. This paper contributes to turnover literature by providing a novel analytical perspective on key elements of turnover models.

J. M. de Oliveira galaxyadvisors AG, Aarau, Switzerland e-mail: [email protected] M. P. Zylka (*) Department of Social Networks & Information Systems, University of Bamberg, Bamberg, Germany e-mail: [email protected] P. A. Gloor MIT Center for Collective Intelligence, Cambridge, MA, USA e-mail: [email protected] T. Joshi Genpact, New York, NY, USA © Springer Nature Switzerland AG 2019 Y. Song et al. (eds.), Collaborative Innovation Networks, Studies on Entrepreneurship, Structural Change and Industrial Dynamics, https://doi.org/10.1007/978-3-030-17238-1_2

43

44

J. M. de Oliveira et al.

1 Introduction Employees or so-called human resources are as important as other resources involved in the production of goods and services. They have a major effect on the productivity of a firm and are responsible for the creation of new knowledge—which may have a significant impact on firm growth. It is not surprising, therefore, that organizations seek to gain benefits by investing in human resources. These benefits, though, diminish when highly skilled employees voluntarily leave an organization for, say, a better position or higher wages at another organization. Further, while losing a highly skilled employee to a competitor may lead to access to external knowledge for the former employer, it also increases the risk that the former employer’s knowledge will be leaked to the competitor (Somaya et al. 2008; Aime et al. 2010). Further, employee turnover reduces human capital of the former employer (Shaw et al. 2005) and is generally found to negatively impact company performance (Hancock et al. 2013). Hence, the competition for highly skilled personnel and their retention is critical to organizations. Such competition was dubbed the “war for talent” (Chambers et al. 1998) back in 1998 and later extended to the “war for Internet talent” (Efrati and Tam 2010), considering developments after the dot-com bubble burst in 2001. Consequently, the turnover of highly skilled employees is a topic of interest for practitioners and scholars alike. It is a well-studied phenomenon in applied psychology, human resource, and general management literature (for an overview of turnover research history, see Holtom et al. 2008). Especially, the causes of actual turnover behavior are focused by research communities, since March and Simon (1958) proposed a first turnover model that takes the reasons for participating in organizations into account. They suggest that employees who are happy with their job and who do not have a job alternative will not leave their employer. Over thousands of scientific articles about employee turnover later, job dissatisfaction became and still is an important proximal cause for employee turnover (Lee and Mitchell 1994). Job dissatisfaction leads to thoughts of quitting and culminates in actual turnover behavior. Hence, it is important to know why and when highly skilled employees become dissatisfied with their current job. To predict such changes and thus turnover behavior, researchers follow a “standard research design” (Steel 2002): data on turnover predictors are collected via a survey, and the information on the actual leaving is collected later. Then, the analysis is conducted with ordinary least squares (OLS) regression, survival, and hazard functions or structural equation modeling (SEM). The accuracy in predicting turnover by this approach is only low to moderate (Lee et al. 2017), slightly higher using turnover intention as a proxy for actual turnover behavior. Novel approaches are necessary that take the trajectories of turnover determinants over time into account because changes over a specific period time in distal precursors affect turnover through changes in proximal antecedents (Hom et al. 2017; Lee et al. 2017). In this paper, we follow the call for novel approaches for employee turnover prediction applying a gated recurrent neural network model (Chung et al. 2014), a

Mirror, Mirror on the Wall, Who Is Leaving of Them All: Predictions for. . .

45

variation of long short-term memory (LSTM) networks (Hochreiter and Schmidhuber 1997; Gers 2001) on longitudinal turnover data. Our goal is to predict turnover of highly skilled employees. Our research questions are the following: • Is a deep learning model suitable for employee turnover prediction? • How does the model perform in the prediction of employee turnover? The remainder of this paper is structured as follows. The next section provides the theoretical grounding. This section contains an overview of related works regarding employee turnover, the role of job (dis-)satisfaction on employee turnover, and the importance of the relational perspective of employee turnover. The subsequent section describes this study’s setting, data, and methodology. The paper concludes with a discussion of the results and implications for future research.

2 Related Work This section begins with a brief overview of employee turnover literature, the role of job satisfaction, and communication behavior of employees on turnover. We then provide an overview of the idea behind recurrent neural networks (RNNs), long short-term memory (LSTM), and gated recurrent units (GRUs), all deep learning architectures that are important for understanding our study’s experimental settings.

2.1

Employee Turnover

The investigation of employee turnover has a long history. March and Simon were the first who define a formal theory of turnover focused on participation in organizations (March and Simon 1958; Lee et al. 2017). Further research focuses on antecedents of individual employee turnover (Mobley 1982). Other research considers external factors such as promotional opportunities and kinship responsibility influencing employee turnover (Price and Mueller 1981). Moreover, studies focus on the negative and positive effects of employee turnover (Dalton and Todor 1979; Mobley 1982). In addition to the obvious negative effect on company performance, turnover can have some other negative effects. These can be differentiated into direct costs (e.g., for recruiting and training the successor) and indirect costs. The latter include loss of firm-specific human capital, demoralization of remaining employees, and loss of social capital embedded in the employees’ relationships (Ton and Huckman 2008). Not surprisingly, loss of social capital from employee turnover negatively impacts company performance (Shaw et al. 2005). But, the opposite can also be the case: employee turnover can potentially lead to the creation of a business tie between the former and new employer, resulting in increased social capital for both firms (Somaya et al. 2008; Corredoira and Rosenkopf 2010).

46

J. M. de Oliveira et al.

More recent research on turnover analyzes the reasons for staying with a company and not the reasons for leaving (Lee et al. 2017). This shift in research objectives is based on the idea of job embeddedness, which incorporates off-the-job and on-thejob factors that embed employees in their job positions (Mitchell et al. 2001). Links to other colleagues, fit within the organization, and potential sacrifices in case of a turnover were found to influence intentions to leave an organization (Mitchell et al. 2001). Especially, relationships with other colleagues and groups are important for high job embeddedness and thus negatively related to turnover intention (Maertz and Griffeth 2004). The analysis of such relations between employees is mostly conducted by a social network analysis approach using centrality measures (c.f. Feeley 2000; Mossholder et al. 2005; Oldroyd and Morris 2012). Brass (1981) found no relationship between being central to an organization’s workflow network and job satisfaction. However, more recent studies show that strong and positive social intraorganizational networks reduce employee turnover (Mossholder et al. 2005; Moynihan and Pandey 2007; Hom and Xiao 2011). Mossholder et al. (2005) found that network centrality might even predict employee turnover. Gloor et al. (2017) analyze the relational view of employee turnover in a detailed manner. Therefore, they analyze e-mail communication of 866 managers with the largest set of network and text analysis metrics so far and show that managers who quit have lower closeness centrality, have less engaged conversations, and shift their communication behavior starting from 5 months before leaving by increasing their degree and closeness centrality, the complexity of their language, as well as their oscillations in betweenness centrality and the number of nudges they need to send to peers before getting an answer. Gloor et al. (2017) show that intraorganizational e-mail communication analyzed with social network and text analysis metrics might be a promising predictor for turnover from the relational perspective. However, the predictive power of past studies of employee turnover that follow the “standard research design” (Steel 2002) is low (Lee et al. 2017). Hom et al. (2017) and Lee et al. (2017) call for research that considers the dynamic nature of antecedents of employee turnover and conduct additional network-based investigations. In the past years, several studies tried different methodologies to predict voluntary employee turnover and to overcome the issue of low predictive power (see Table 1 for review results). Especially, machine learning-based classification algorithms like random forests (RF), gradient boosting machines (GBM), support vector machines (SVM), naïve Bayes (NB), and k-nearest neighbors (kNN) are applied to employee turnover data typically gathered from human resource information systems (HRIS) or questionnaire surveys. Predictors are mostly sociodemographic and work-related items like working experience, tenure, skills, and performance. These studies present good accuracy values and claim that the machine learning-based classification algorithms outperform logit models (e.g., Nagadevara et al. 2008; Punnoose and Ajit 2016; Zhao et al. 2018). Our study follows the call of Lee et al. (2017) and is influenced by the study Gloor et al. (2017). However, we want to go one step further in turnover prediction by using recurrent neural networks (RNNs) that might enhance the predictive power of the relational perspective of employee turnover.

Workers of a small Midwest manufacturing company, USA

Sexton et al. (2005)

Ribes et al. (2017)

Leadership team of a global retailer, USA Workers of a regional child welfare agency, USA Managers

Study context Employees in manufactory industries Marketing specialists of motor marketing company, Taiwan Software company, India

Punnoose and Ajit (2016) Quinn et al. (2002)

Nagadevara et al. (2008)

Hong et al. (2005)

Research Chang (2009)

Voluntary turnover

Reviewing personnel files

Not reported

447

1000

536

Provided by internal IS

Turnover behavior

Voluntary turnover

73,115

150

132

Sample size 881

Provided by HRIS

Secondary data provided by the company

Not specified

Data source Not specified

Voluntary turnover

Voluntary turnover

Voluntary turnover

Turnover context Voluntary turnover

Neural network simultaneous optimization algorithm

NB, LDA, SVM, RF

Neural network, logistic regression, classification and regression trees, classification trees (C5.0), discriminant analysis Logit model, NB, RF, kNN, LDA, SVM, XGBoost Logit model, multilayered perceptron

Methodology Taguchi method + nearest neighbor classification rules SVM, logistic regression, probit model

Demographic factors, skill set, working experience Demographic factors, skill set, working experience, job level

Demographic factors, skill set, working experience Demographic factors, skill set, working experience

Demographic factors, skill set, working experience

Predictor Demographic factors, skill set, working experience Constant job performance

Table 1 An overview of the literature on the application of machine learning algorithms in the context of employee turnover

(continued)

Accuracy 0.96 with neural network

ROC ¼ 0.95 with Random Forest

Accuracy ¼ 0.60 (test set) with multilayered perceptron

AUC ¼ 0.86 with XGBoost

Accuracy ¼ 0.85 with C5.0

Accuracy ¼ 0.84 with SVM

Best model performance Accuracy ¼ 0.878%

Mirror, Mirror on the Wall, Who Is Leaving of Them All: Predictions for. . . 47

Study context Workers of an automotive parts manufacturing company, Iran

Hospital employees, USA

Leading IT organization

Nurses, Taiwan

Employee of a regional bank, USA

Research Sikaroudi et al. (2015)

Somers (1999)

Suceendran et al. (2015)

Tzeng et al. (2004)

Zhao et al. (2018)

Table 1 (continued)

Turnover behavior

Turnover intention

Attrition

Voluntary turnover

Turnover context Voluntary turnover

Dataset 1: not reported Dataset 2: IBM Watson database

Exit interview details of employees Questionnaire survey

Questionnaire survey

Data source Provided by HRIS

Dataset 1: 9089 Dataset 2: 1470

389

2572

577

Sample size Not reported

Decision tree, RF, GBM, XGBoost, logit model, SVM, neural networks, LDA, NB, kNN

ID3, J48, NB, Bayesian network, K star, IBK, random tree, RF SVM

Methodology Multilayer perceptron, RF, probabilistic neural network, SVM, classification and regression tree, kNN, NB, Apriori, CN2 algorithm Multilayer perceptron, learning vector quantization Affective commitment, continuance commitment, job satisfaction, job withdrawal intentions Demographic factors, skill set, working experience, performance Working motivation, job satisfaction, and stress levels Demographic factors, skill set, working experience, performance

Predictor Demographic factors, skill set, working experience

ROC ¼ 0.9008 with Extreme Gradient Boosting

Accuracy ¼ 0.97 with IBK and random tree Accuracy ¼ 0.892 with SVM

Accuracy ¼ 0.88 with multilayer perceptron

Best model performance Accuracy ¼ 0.906 with Random Forest

48 J. M. de Oliveira et al.

Mirror, Mirror on the Wall, Who Is Leaving of Them All: Predictions for. . .

2.2

49

Recurrent Neural Networks (RNN) and Gated Recurrent Units (GRU)

Recurrent neural networks (RNNs) are artificial neural networks that encode knowledge dependencies learned over past events and use this knowledge to reason about current events. RNNs achieve this by using cycles in a network, which allows past knowledge to persist in the form of inputs to the next network. RNNs can be thought of as a directed graph where each node passes the knowledge on to the next node after applying set of weights and transformations. Hence, RNNs are appropriate for data in the form of sequence or data with temporal attributes. In recent years, RNNs have been used successfully in a variety of contexts, like object detection (Szegedy et al. 2013), speech recognition (Graves et al. 2013), or classification problems (Krizhevsky et al. 2012). Standard RNNs have a few limitations. One of them is that they cannot process inputs with varying length because long-term knowledge of too long input sequences cannot be stored (vanishing gradient problem). To process long-range sequences and identify relevant reasons for employee turnover, we need a model that can remember long-term knowledge. Long short-term memory (LSTM) is an RNN architecture that avoids the vanishing gradient problem (Hochreiter and Schmidhuber 1997; Gers 2001) and learns tasks that require knowledge of events that happened lots of time series earlier (Schmidhuber 2015). Therefore, we use a LSTM architecture with a simpler gating mechanism called gated recurrent units (GRU). GRUs are comparable to LSTM in terms of performance and exhibit better performance on smaller datasets (Chung et al. 2014).

3 Experiments In this section, we present a brief description of our data, describe the data preprocessing steps, continue with the experimental settings, and conclude this section by presenting the results of our experiments.

3.1

Data Preprocessing

This study is based on employees’ e-mail communication data that is provided by a global professional services firm with more than 70,000 employees in over 20 countries. The e-mail data contain 845,208 actors (employees1 and external stakeholders), ~138 million edges representing the communication that took place from 1 January 2017 till 29 January 2018. 1 Some employees had more than one e-mail account in this company. In these cases, we merged the multiple e-mail accounts of an employee into one.

50

J. M. de Oliveira et al.

We filtered the 845,208 actors by employees in managerial positions, which led to e-mail data of 3952 managers containing 35% (48 million of 138 million edges) of the provided e-mail communication data. After collecting the mailboxes of the 3952 managers, we prepared the data for the deep learning experiments. Therefore, we split the data in time frames (15 days time window with 3 days separation for the next window) and calculated network metrics (see Table 2) for each time frame with Condor,2 a social network and text analysis software. Each time frame was exported in a separate CSV file. The calculated network metrics are capable to show the changes in communication patterns of employees (c.f. Gloor et al. 2017) and hence serve as input variables for the deep learning models. For the machine learning experiments, we consider each CSV file as a single data point, meaning we combine 30 CSV files to have network metrics related to 3 months of e-mail communication as the input. Depending on the date an employee left the company, we consider her as a leaver or not. The last 3 months of work of employees who left the company are not considered in the model, because employees in this company are asked to send the resignation letter 3 months before leaving. Following Gloor et al. (2017), we assume that the closer employees get to the final decision of quitting, the higher the likelihood to exhibit divergent communication behaviors. Hence, we define 5–7 months prior turnover as the period, where an employee is thinking about leaving the company. This means that 150 days before the actual turnover job satisfaction might turn to job dissatisfaction (see Figs. 1 and 2).

3.2

Experiment Setup

We built an RNN with GRU. We trained our model on 2898 employees and tested on the remaining 954 employees. Further, we apply fourfold cross-validation to provide a robust estimate of the performance of our model. Therefore, we split the training data into four equal-sized subsets; each subset has a similar number of employees who left (leavers). One of the four subsets is retained as the validation data for testing, and the remaining three subsets are used for training. Since our dataset contains only a small number of actual employees who left the company (78 leavers), we had to preprocess the data with synthetic minority over-sampling technique (SMOTE) to handle the imbalanced dataset (see Table 3 for dataset statistics). SMOTE is an over-sampling approach in which the minority class, in our case the leavers, is over-sampled. But instead of over-sampling with replacements, synthetic instances are created by joining any of the k minority class nearest neighbors of a minority class. These synthetic instances lead to more sensitivity regarding a minority class without too severe minority over-sampling. We evaluate the results using precision, recall, accuracy, area under curve (AUC), and Matthews correlation coefficient (MCC) score, a discretization of the Pearson

2

http://guardian.galaxyadvisors.com

Mirror, Mirror on the Wall, Who Is Leaving of Them All: Predictions for. . .

51

Table 2 Calculated network metrics that are considered as input variables for the machine learning models Metric Messages sent Messages received Messages total Words total Degree centrality Betweenness centrality

Betweenness centrality oscillation Closeness centrality

Reach 2 Contribution index

Contribution index oscillation Ego ART

Ego nudges Alter ART

Alter nudges Avg. influence per message Total influence

Avg. sentiment

Avg. emotionality Avg. complexity

Definition # of e-mails sent by an employee # of e-mails received by an employee Sum of e-mails sent and received by an employee # unique words an employee used in her e-mails # of colleagues each employee is directly connected within the communication network (Freeman 1978) Likelihood to be on the shortest path between any two actors in the network. Indicates the extent to which each employee acts as an information hub and controls the information flow (Freeman 1978) # of local maxima and minima in the betweenness curve of an actor. Indicates how frequently employees change their network position in the team, from central to peripheral and back Inverse of distance of an actor from all other actors in the network, considering the shortest paths that connect each pair of actors. Indicates the efficiency of transmitting information and independence from other peripheral actors (Friedkin 1991) # of colleagues each employee can reach by the distance of two # messages sent—# messages received/(# messages sent + # messages received). Indicates how balanced a communication is in terms of sent and received messages # of local maxima and minima in the contribution index curve of an actor Avg. # of hours sender takes to respond to e-mails. Time until a frame is closed for the receiver after she has sent an e-mail. Indicates the respect the receiver has for the sender Avg. # of follow-ups that the sender needs to send to receive a response from the receiver Avg. # of hours receiver takes to respond to e-mails. Time until a frame is closed for the sender, after she has sent an e-mail. Indicates the respect the receiver has for the sender Avg. # of follow-ups that the receiver needs to send to receive a response from the sender Avg. # of terms per message that has been introduced into the network # of new terms which a sender has introduced into the network and which are subsequently used by other members of the network. Indicates the extent to which someone causes the other person’s pattern of speaking to match their own pattern Uses automatically generated bag of word, based on a dictionary trained for language/subject area. Indicates positivity and negativity of communication Standard deviation of sentiment. It represents the deviation from neutral sentiment Information distribution using TF/IDF, independent of single words. Indicates the complexity of word usage. The more diverse words, which are all used evenly, a sender uses, the higher the complexity

52

J. M. de Oliveira et al.

Fig. 1 Employee satisfaction timeline

correlation value. We chose MMC as an additional performance measure because our model represents a binary classification problem (leavers/stayers) and MCC is more informative than other confusion matrix measures (such as F1 score and accuracy) in evaluating binary classification problems. MCC takes the balance ratios of the four confusion matrix categories (true positives, true negatives, false positives, false negatives) into account (Chicco 2017).

3.3

Results

Tables 4 (training set) and 5 (test set) show fourfold cross-validation results of our dataset. We conducted the GRU experiment with a keep probability of 0.1, one LSTM layer, eight neurons, and a learning rate of 0.001. All folds perform well; the average fold has an ACC of 0.933 and an MCC of 0.873. The test set’s performance is good. The ACC is 0.800, and MCC (0.615) shows a strong positive predictive power (Figs. 3 and 4). Table 6 shows the performance of the two GRUs with different configurations. Model 2 with 16 neurons performs slightly better (MCC ¼ 0.615) than Model 1 with 8 neurons (MCC ¼ 0.554).

4 Discussion and Implications The results of the experiments revealed that applying a deep learning approach has potential to conduct a binary classification of employees in stayers or leavers by analyzing their e-mail communication behavior.

Mirror, Mirror on the Wall, Who Is Leaving of Them All: Predictions for. . . Alter Nudges

0.35

53

Ego Nudges 0.40

0.30

0.35

0.25

0.30

0.20

0.25

0.15

0.20 0.15

0.10

0.10

0.05

0.05

0.00

17–08

17–09

17–08

17–09

17–08

17–09

17–07 17–07

17–06

17–05

17–04

17–07

Degree centrality

17–03

17–02

17–01

17–09

17–08

17–07

17–06

17–05

17–04

17–03

17–02

17–01

0.00

Closeness centrality 0.96

0.6

0.94

0.5

0.92

0.4

0.90

0.3

0.88 0.86

0.2

17–06

17–05

17–04

17–03

17–01

17–09

17–08

17–07

17–06

17–05

17–04

17–03

17–02

0.80 17–01

0.82

0.0

17–02

0.84

0.1

Reach 2

Betweenness centrality 0.8

0.12 0.10

0.7

0.08 0.6

0.06 0.04

0.5

0.02 17–06

17–05

17–04

17–03

17–09 17–09

17–02

17–08

17–07

17–08

17–06

17–05

17–04

17–03

17–02

17–01

17–01

0.4

0.00

avg complexity 0.8 0.6 0.4 0.2

17–07

17–06

17–05

17–04

17–03

17–02

17–01

0.0

Fig. 2 An exemplary illustration of selected metrics of an employee. The x-axis indicates the time (monthly scale), and the y-axis indicates the values of the metric. The three phases of the employee satisfaction timeline are color-coded in these charts. Blue indicates the period in which the

54 Table 3 Statistics of the dataset

Table 4 Model results, training set

J. M. de Oliveira et al.

# of actors # of leavers # of stayers’ time series # of leavers’ time series # of stayers’ SMOTE time series # of leavers’ SMOTE time series

Fold # 0 1 2 3 1.5 (avg.)

ACC 0.949 0.960 0.911 0.911 0.933

P 0.997 0.995 0.999 1000 0.998

Training 2898 59 238,438 832 238,438 238,438

R 0.911 0.930 0.850 0.853 0.886

AUC 0.954 0.962 0.924 0.926 0.942

Test 954 19 78,468 262 78,468 78,468

MCC 0.901 0.921 0.835 0.835 0.873

ACC accuracy, P precision, R recall, AUC area under curve, MCC Matthews correlation coefficient Table 5 Model results, test set

Fold # 0 1 2 3 1.5 (avg.)

ACC 0.796 0.740 0.883 0.780 0.800

P 0.700 0.547 0.962 0.770 0.745

R 0.867 0.891 0.830 0.786 0.844

AUC 0.808 0.782 0.893 0.780 0.816

MCC 0.604 0.520 0.775 0.561 0.615

ACC accuracy, P precision, R recall, AUC area under curve, MCC Matthews correlation coefficient

What are the implications from this study’s findings for future RNN experiments and research on employee turnover? First, our study provides a novel approach to analyze longitudinal employee turnover data. This study is among the first to apply an RNN besides the usual applications of RNNs, such as object or speech recognition. As such, this paper may provide an insightful account for researchers interested in the context of employee turnover and provide an example of how deep learning methodology can be meaningful integrated in management research studies in general. We found it encouraging that the experiments went well in regard of performance. Further, we address the call by Hom et al. (2017) as well as Lee et al. (2017) by considering the dynamic nature of antecedents of employee turnover and conducting a network-based analysis in comparison to earlier research that primarily used a standard research approach (Steel 2002).

Fig. 2 (continued) employee is satisfied. Yellow shows the phase in which the employee thinks about leaving the company. Red indicates the phase in which the employee has decided to leave the company and has signed the resignation letter

Mirror, Mirror on the Wall, Who Is Leaving of Them All: Predictions for. . .

1.00 0.95 0.90 0.85 0.80 0.75 0.70 0.65 0.60 0.55 0.50

2000

8000 ACC

P

14000 R

55

18000

AUC

MCC

Fig. 3 Development of model performance indicators by model steps, training set, avg. fold

1.00 0.90 0.80 0.70 0.60 0.50 0.40

2000

8000 ACC

P

14000 R

18000

AUC

MCC

Fig. 4 Development of model performance indicators by model steps, test set, avg. fold

Table 6 Model results based on different configuration (test set, 1.5-fold) Model 1

Model 2

Configuration parameters Keep probability: 0.1 Layers: 1 Size: 8 Learning rate: 0.001 Type: “GRU” Keep probability: 0.1 Layers: 1 Size: 16 Learning rate: 0.001 Type: “GRU”

ACC 0.768

P 0.653

R 0.849

AUC 0.786

MCC 0.554

0.800

0.745

0.844

0.816

0.615

56

J. M. de Oliveira et al.

For organizations, the findings suggest the critical importance of human resource (HR) data analytics. We provide a possibility to predict employee turnover. HR managers can use our experiment setup as an “early warning system” for employee turnover. Since employee turnover might be dysfunctional and has serious impact on company performance, HR managers could counteract with retention strategies when an important and highly skilled employee intend to leave the company. Nevertheless, ethical issues should be considered before applying this methodology in an organization.

5 Limitations and Future Research One limitation of this work pertains to the generalizability of our proposed approach. It is plausible that the insights from this study might not directly apply to other companies or occupational groups. However, that would be a practical concern caused by insufficient data, which should be manageable. Second, the premise of this study is that dissatisfaction derived from e-mail communication behavior culminates in leaving. Other options or paths of dissatisfied employees are ignored. Employees may lower job inputs or improve their circumstances (via promotion) rather than leave (with or without job offers) (Hulin et al. 1985). Thus, leaving is only one option among many ways to cope with job dissatisfaction. Third, employee turnover data is an imbalanced dataset by nature. The number of leavers is always much lower than the number of stayers. Our dataset is extremely imbalanced, but we overcome this issue with SMOTE. However, a higher sample of leavers might improve the predictive performance of our models. Fourth, this paper takes the relational perspective on employee turnover. However, traditional antecedents like organizational commitment were not considered. Future research should conduct a classifier performance comparison by including several configurations of LSTMs, GRUs, and other classifier models like support vector machines (SVMs). Additionally, a comparison with models that are based on the mentioned standard research approach is necessary. Further, this study should be replicated with employee turnover data from other organizations, other occupational groups and with additional input variables.

6 Conclusion In this paper, we applied a GRU RNN classifier that classifies employees in two states (leaver or stayer) by taking their e-mail communication behavior into account. The classifier’s performance is measured in terms of confusion matrix with accuracy, recall, precision, AUC, and MCC values. The developed GRU RNN model provides promising performance. Here, GRU can strongly benefit from the fact that it can look back in time and learn to correlate the calculated network metrics. GRU can learn

Mirror, Mirror on the Wall, Who Is Leaving of Them All: Predictions for. . .

57

these correlations, although it might require further training or different variables to be added to the data. In the future, we will continue to improve the performance of our model and conduct in-depth error analysis. We finally conclude that GRU is very suitable for classifying employees’ turnover behavior. This is the first reported demonstration of a successful application of GRU neural networks to data from an organizational management context, namely, employee turnover.

References Aime, F., Johnson, S., Ridge, J. W., & Hill, A. D. (2010). The routine may be stable but the advantage is not: Competitive implications of key employee mobility. Strategic Management Journal, 31, 75–87. https://doi.org/10.1002/smj.809 Brass, D. J. (1981). Structural relationships, job characteristics, and worker satisfaction and performance. Administrative Science Quarterly, 26, 331. https://doi.org/10.2307/2392511 Chambers, E., Foulon, M., Handfield-Jones, H., Hanking, S. M., & Michaels, E. G., III. (1998). War for talent. The McKinsey Quarterly, 3, 44–57. Chang, H. Y. (2009). Employee turnover: A novel prediction solution with effective feature selection. WSEAS Transactions on Information Science and Applications, 6, 417–426. Chicco, D. (2017). Ten quick tips for machine learning in computational biology. BioData Mining, 10, 35. https://doi.org/10.1186/s13040-017-0155-3 Chung, J., Gulcehre, C., Cho, K., & Bengio, Y. (2014). Empirical evaluation of gated recurrent neural networks on sequence modeling. Neural Information Processing Systems 2014 Deep Learning and Representation Learning Workshop. Corredoira, R. A., & Rosenkopf, L. (2010). Should auld acquaintance be forgot? The reverse transfer of knowledge through mobility ties. Strategic Management Journal, 31, 159–181. https://doi.org/10.1002/smj.803 Dalton, D. R., & Todor, W. D. (1979). Turnover turned over: An expanded and positive perspective. Academy of Management Review, 4, 225–235. Efrati, A., & Tam, P.-W. (2010). Google battles to keep talent. Wall Street Journal. Accessed May 5, 2014, from http://www.wsj.com/articles/SB10001424052748704804504575606871487743724 Feeley, T. H. (2000). Testing a communication network model of employee turnover based on centrality. Journal of Applied Communication Research, 28, 262–277. https://doi.org/10.1080/ 00909880009365574 Freeman, L. C. (1978). Centrality in social networks conceptual clarification. Social Networks, 3, 215–239. Gers, F. (2001). Long short-term memory in recurrent neural networks (Vol. 2366, p. 102). Lausanne: EPFL. https://doi.org/10.1162/neco.1997.9.8.1735 Gloor, P. A., Fronzetti Colladon, A., Grippa, F., & Giacomelli, G. (2017). Forecasting managerial turnover through e-mail based social network analysis. Computers in Human Behaviour, 71, 343–352. https://doi.org/10.1016/j.chb.2017.02.017 Graves, A., Mohamed, A., & Hinton, G. (2013). Speech recognition with deep recurrent neural networks. In 2013 IEEE international conference on acoustics, speech and signal processing (pp. 6645–6649). IEEE Hancock, J. I., Allen, D. G., Bosco, F. A., McDaniel, K. R., & Pierce, C. A. (2013). Meta-analytic review of employee turnover as a predictor of firm performance. Journal of Management, 39, 573–603. https://doi.org/10.1177/0149206311424943 Hochreiter, S., & Schmidhuber, J. (1997). Long short-term memory. Neural Computation, 9, 1–32. https://doi.org/10.1162/neco.1997.9.8.1735

58

J. M. de Oliveira et al.

Holtom, B. C., Mitchell, T. R., Lee, T. W., & Eberly, M. B. (2008). 5 Turnover and retention research: A glance at the past, a closer review of the present, and a venture into the future. Academy of Management Annals, 2, 231–274. https://doi.org/10.1080/19416520802211552 Hom, P. W., Lee, T. W., Shaw, J. D., & Hausknecht, J. P. (2017). One hundred years of employee turnover theory and research. The Journal of Applied Psychology, 102, 530. Hom, P. W., & Xiao, Z. (2011). Embedding social networks: How Guanxi ties reinforce Chinese employees’ retention. Organizational Behavior and Human Decision Processes, 116, 188–202. https://doi.org/10.1016/j.obhdp.2011.06.001 Hong, W. C., Pai, P. F., Huang, Y. Y., & Yang, S. L. (2005). Application of support vector machines in predicting employee turnover based on job performance. Advanced Natural Computation, LNCS, 3610, 668–674. Hulin, C. L., Roznowski, M., & Hachiya, D. (1985). Alternative opportunities and withdrawal decisions: Empirical and theoretical discrepancies and an integration. Psychological Bulletin, 97, 233–250. https://doi.org/10.1037/0033-2909.97.2.233 Krizhevsky, A., Sutskever, I., & Hinton, G. (2012). ImageNet classification with deep convolutional neural networks. Advances in Neural Information Processing Systems, 25, 1097–1105. Lee, T. W., Hom, P. W., Eberly, M. B., & Mitchell, T. R. (2017). On the next decade of research in voluntary employee turnover. Academy of Management Perspectives, 31, 201–221. https://doi. org/10.5465/amp.2016.0123 Lee, T. W., & Mitchell, T. R. (1994). An alternative approach: The unfolding model of voluntary employee turnover. Academy of Management Review, 19, 51–89. https://doi.org/10.5465/AMR. 1994.9410122008 Maertz, C. P., & Griffeth, R. W. (2004). Eight motivational forces and voluntary turnover: A theoretical synthesis with implications for research. Journal of Management, 30, 667–683. https://doi.org/10.1016/j.jm.2004.04.001 March, J. G., & Simon, H. A. (1958). Organizations. Cambridge, MA: Wiley-Blackwell. Mitchell, T. R., Holtom, B. C., Lee, T. W., Sablynski, C. J., & Erez, M. (2001). Why people stay: Using job embeddedness to predict voluntary turnover. Academy of Management Journal, 44, 1102–1121. https://doi.org/10.2307/3069391 Mobley, W. H. (1982). Some unanswered questions in turnover and withdrawal research. Academy of Management Review, 7, 111–116. Mossholder, K. W., Settoon, R. P., & Henagan, S. C. (2005). A relational perspective on turnover: Examining structural, attitudinal, and behavioral predictors. Academy of Management Journal, 48, 607–618. https://doi.org/10.5465/amj.2005.17843941 Moynihan, D. P., & Pandey, S. K. (2007). The ties that bind: Social networks, person-organization value fit, and turnover intention. Journal of Public Administration Research and Theory, 18, 205–227. https://doi.org/10.1093/jopart/mum013 Nagadevara, V., Srinivasan, V., & Valk, R. (2008). Establishing a link between employee turnover and withdrawal behaviours: Application of data mining techniques. Research and Practice in Human Resource Management, 16, 1–27. Oldroyd, J. B., & Morris, S. S. (2012). Catching falling stars: A human resource response to social capital’s detrimental effect of information overload on star employees. Academy of Management Review, 37, 396–418. https://doi.org/10.5465/amr.2010.0403 Price, J. L., & Mueller, C. W. (1981). A causal model of turnover for nurses. A causal model of turnover for nurses. Academy of Management Journal, 24, 543–565. https://doi.org/10.2307/ 255574 Punnoose, R., & Ajit, P. (2016). Prediction of employee turnover in organizations using machine learning algorithms. International Journal of Advanced Research in Artificial Intelligence, 9, 22–26. Quinn, A., Rycraft, J. R., & Schoech, D. (2002). Building a model to predict caseworker and supervisor turnover using a neural network and logistic regression. Journal of Technology in Human Services, 19, 65–85. Ribes, E., Touahri, K. & Perthame, B. (2017). Employee turnover prediction and retention policies design: A case study. arXiv preprint:1707.01377.

Mirror, Mirror on the Wall, Who Is Leaving of Them All: Predictions for. . .

59

Schmidhuber, J. (2015). Deep learning in neural networks: An overview. Neural Networks, 61, 85–117. https://doi.org/10.1016/j.neunet.2014.09.003 Sexton, R. S., McMurtrey, S., Michalopoulos, J. O., & Smith, A. M. (2005). Employee turnover: A neural network solution. Computers and Operations Research, 32, 2635–2651. Shaw, J. D., Duffy, M. K., Johnson, J. L., & Lockhart, D. E. (2005). Turnover, social capital losses, and performance. Academy of Management Journal, 48, 594–606. https://doi.org/10.5465/ AMJ.2005.17843940 Sikaroudi, A. M. E., Ghousi, R., & Sikaroudi, A. E. (2015). A data mining approach to employee turnover prediction (Case study: Arak automotive parts manufacturing). Journal of Industrial and Systems Engineering, 8, 106–121. Somaya, D., Williamson, I. O., & Lorinkova, N. (2008). Gone but not lost: The different performance impacts of employee mobility between cooperators versus competitors. Academy of Management Journal, 51, 936–953. https://doi.org/10.5465/AMJ.2008.34789660 Somers, M. J. (1999). Application of two neural network paradigms to the study of voluntary employee turnover. Journal of Applied Psychology, 84(2), 177. Steel, P. (2002). At the empirical turnover theory of fit interface: Problems. Academy of Management Review, 27, 346–360. https://doi.org/10.5465/amr.2002.7389900 Suceendran, K. M., Saravanan, R., Ananthram, D., Poonkuzhali, S., Kumar, R. K., & Sarukesi, K. (2015). Applying classifier algorithms to organizational memory to build an attrition predictor model. Szegedy, C., Toshev, A., & Erhan, D. (2013). Deep neural networks for object detection. In Neural information processing systems conference (pp. 2553–2561). Ton, Z., & Huckman, R. S. (2008). Managing the impact of employee turnover on performance: The role of process conformance. Organization Science, 19, 56–68. https://doi.org/10.1287/ orsc.1070.0294 Tzeng, H. M., Hsieh, J. G., & Lin, Y. L. (2004). Predicting nurses’ intention to quit with a support vector machine: A new approach to set up an early warning mechanism in human resource management. Computers, Informatics, Nursing, 22, 232–242. Zhao, Y., Hryniewicki, M. K., Cheng, F., Fu, B., & Zhu, X. (2018). Employee turnover prediction with machine learning: A reliable approach. In Proceedings of SAI intelligent systems conference (pp. 737–758). Cham: Springer.

Education and Technology as Levers for Sustainable Change A New Framework of Interaction Between Business and Environment Carlo Alberto Amadei and Monica Baraldi Borgida

Abstract In the last decade, researchers have increasingly analyzed the interaction between business and environment under new perspectives. We join this effort focusing on the positive change that can arise from new collaborative opportunities between business and the environment, in order to embrace environmental challenges and pursue reciprocal benefits. We evaluate this synergetic activity using system thinking which identifies pivotal center, has the power to originate connections, and regulates a system’s behavior. In this regard, we believe education and technology are among the most important elements in the system. They play a major role in determining and weighing the interactions between environment and business. The two scenarios presented here are (1) the status quo system and (2) a system in which technology and education are empowered and serve as leverage points to develop a more efficient use of the world resources, thus using a smaller portion of the world’s physical capacity. This last scenario focuses on one environmental sphere (water), but the same tool could be applied to other environmental aspects. We build our analysis balancing the contributions coming from both professional and academic spheres. Progress has been made to integrate sustainability into strategy, such as the adoption of science-based targets by organizations aiming to reduce the effect of climate change. The authors believe that this process explains new ways in which business and society can thrive for generations to come.

C. A. Amadei Harvard John A. Paulson School of Engineering and Applied Sciences, Cambridge, MA, USA M. B. Borgida (*) Northeastern University, Boston, MA, USA e-mail: [email protected] © Springer Nature Switzerland AG 2019 Y. Song et al. (eds.), Collaborative Innovation Networks, Studies on Entrepreneurship, Structural Change and Industrial Dynamics, https://doi.org/10.1007/978-3-030-17238-1_3

61

62

C. A. Amadei and M. B. Borgida

1 Introduction The idea for this paper is the product of an interdisciplinary conversation across the fields of business and environmental engineering. When we met and shared our professional interests, it was clear that there was an overlapping of activity in our research questions, our concerns, and especially the solutions that we had in mind. We started our journey looking at the big picture, thinking of the possibilities for business and environment to contribute to more sustainable global living conditions (Lovelock 2009). Since both business and engineering focus on satisfying human needs in a sustainable way, we looked at the population trends and the resources available to support the current consumption system. We found that, in 2017, there were 7.2 billion people and the world population continues to rise rapidly. Several sources anticipate that there will be 8 billion people by the 2020s on the way to become 9 billion by the early 2040s (Sachs 2015). Improved education and health-care systems will possibly affect this dynamic, but it is difficult to forecast a different trend in the immediate future. The world population is looking for economic improvement, both in poor and rich areas, in a world that is more and more interconnected through trade, finance, technologies, supply chains, migration, and social networks (Matson et al. 2016). However, this growth process, both on the side of the population and on the side of the economy, is often unbalanced (Kolbert 2015). Locally and globally, we find evidence of a remarkable environmental crisis, and the pursuit of economic growth is one of its main drivers (Henderson et al. 2015). We need to think and adopt solutions that change the entire system of production, distribution, and consumption; we need to shift from a state of mind shaped around the concept of “surplus” to a new systemic view based on the condition of “scarcity” for a more sustainable human civilization (Mullainathan and Shafir 2014). This condition of “scarcity” is the focus point of the analysis involved in our theoretical model that frames a new relationship between business and environment. With regard to sustainability, we adopted the view designed within the 2016 Global Sustainable Development Report, with special reference to the economic, social, and environmental objectives and political systems of the Sustainable Development Goals (SDGs) (UN 2015). We also looked at the sustainability trends from a historical perspective (Caradonna 2014). We join this effort focusing on the positive change that can arise from new collaborative opportunities between business and the environment, in order to embrace environmental challenges and pursue reciprocal benefits. Specifically, we evaluate this synergetic activity using system thinking which identifies pivotal center, has the power to originate connections, and regulates the system’s behavior. In this regard, we believe education and technology are among the most important elements which play a major role in determining and weighing the interactions between environment and business. The two scenarios presented in this paper are (1) the status quo system and (2) a system in which technology and education are

Education and Technology as Levers for Sustainable Change

63

empowered and serve as leverage points to develop a more efficient use of the world resources, thus using a smaller portion of the world’s physical capacity. This last scenario focuses on one environmental sphere (water), but the same tool could be applied to other environmental aspects. Fundamentally, we aim to design a system of relationships that create a win-win model rather than support a paradigm that favors one actor only (business) at the expense of the other (environment). For our purposes, we choose to envision positive and negative loops that firms, public institutions, NGOs, educational organizations, and public awareness play in the interaction between business and environment. We hope that the resulting system that we describe will inspire further research, improve business practices, and promote institutional reforms that embrace collaboration and synergies rather than isolated actions.

2 The Vulnerability of the Current System: A Possible Way Out According to the Brundtland Report issued in 1987 by the United Nations World Commission on Environment and Development (WCED), “Sustainable Development is development that meets the needs of the present without compromising the ability of future generations to meet their own needs” (Commission 1987). Similarly, other sources show increasing alignment around the concept that achieving prosperity in its fullest meaning does not imply to expand the economy indefinitely. In pursuing the UN Sustainable Development Goals, it is acceptable to move public policy objectives away from the pursuit of economic growth and toward specific targets linked to the well-being of humans and other species. In line with some of the most relevant research in the field, we recognize that the normative side of sustainability envisions three basic objectives of a good society: economic prosperity, social inclusion, and environmental sustainability (Tirole 2017). What we have witnessed is that, over time, this normative and ethical concept of sustainable development has become more and more focused on the design of complex systems that require interdisciplinary thinking applied to the abovementioned areas: • • • •

Economic development Social inclusion Environmental sustainability Good governance for all the major social actors, including government and business, which should facilitate the interaction in the abovementioned three spheres

Our research methodology, in terms of approach, looked at the holistic view adopted by the abovementioned UN report (Commission 1987); we specifically focused our attention on two dimensions, business and environment, and noticed two trends.

64

C. A. Amadei and M. B. Borgida

First, with regard to the environment, there is evidence that the global economic activity is changing the earth’s climate and its main cycles related to water, soil, and air (Tubiana 2018); human activities disrupt ecological systems worldwide and carry a higher risk of spreading disease among animals and humans (Kolbert 2015). Moreover, overfishing is compromising the oceans’ ability to supply the protein that sustains billions of people. Loss or degradation of key agricultural resources is leading to a global agricultural system that depends on international markets for basic food supplies (Hawken 2017). Global warming increases ocean levels around the world; “over the past 40 years, the upper 75 meters of the world’s oceans have warmed at an average rate of more than 0.1 degrees Celsius” (Auth 2015). Overall, the scale of impact carried by the world economy is creating major disruptions of several earth’s ecosystems. Species are going extinct, and we witness the effects of climate change (Sachs 2014). Second, with reference to business, there is evidence that for many firms sustainability is not a full opportunity yet. According to a global study conducted from 2009 to 2017 in 118 countries and with 60,000 respondents, 90% of companies consider a sustainability strategy important to remaining competitive, 60% have a sustainability strategy, and only 25% have a business case for sustainability (Sloan/BCG 2017). The same research explains this result, reporting that the majority of business leaders tend to view sustainable factors as necessary requirements, rather than as sources of business opportunity. Among the incentives to report on sustainable initiatives, managers mention increase reputation and brand building and improve relationships with employees and other stakeholders (Ernst & Young & Boston College Center for Corporate Citizenship 2013). In contrast, companies with embedded sustainability strategies view the relationship between business and environment as an opportunity to innovate their products and business processes rather than as a constraint and show a better financial performance in the long term (Eccles et al. 2014). These organizations are also defined as purpose-driven companies, whose goal is to internalize sustainable needs and apply new frameworks to refine the company strategy. Among the main issues that drive embedded sustainability in business, we can list corruption, labor conditions, human rights, resource scarcity, water access, and climate change (Sheffi 2018). Examples of best practices that are emerging in the field of sustainability and imply relevant engagement on the business side are: • A new generation of Global Reporting Initiative guidelines (G4): GRI is an international independent organization based in Amsterdam that helps businesses, governments, and other organizations understand and communicate the impact of their business on critical sustainability issues. The first GRI standards useful to provide the abovementioned feedback were issued in 1997; in 2018 the GRI reports over 9000 reporting organizations. Of the world’s largest 250 corporations, 93% report on their sustainability performance, and 82% of these adopt the GRI’s standards (GRI 2018a, b).

Education and Technology as Levers for Sustainable Change

65

• Sustainable Accounting Standards (SAS): This initiative has been a major innovation in the accounting field and within the current system of financial regulation. The Sustainable Accounting Standard Board (SASB), a San Francisco-based organization, aims to integrate its sustainability accounting standards into the Form 10-K which must be filed by public companies with the US Securities and Exchange Commission. As of 2018, SASB has issued sustainability accounting standards in 77 industries organized in major sectors like consumption, finance, health care, infrastructure, nonrenewable resources, renewable resources and alternative energy standards, services, technology and communications, and transportation (SASB 2018). • Increasing initiatives of industry autoregulation for sustainable purposes. Research shows that models of environmental public policy can inspire different patterns of environmental adaptation by firms. Voluntary policies, and within them those based on cooperation, can become the most stimulating for the adoption of proactive environmental practices, especially those more innovative practices that extend beyond legal regulation. Moreover, research confirms that the adoption of more advanced environmental innovations by firms submitted to autoregulation allowed both a superior improvement and a higher level of environmental performance (Sheffi 2018). It has been the case of companies like Unilever, Natura, Siemens Sustainable Lobster Fishing, and Valeo. These firms have been at the forefront of sustainability trends started by the company’s initiative (Knights 2018). • Increasing shared value initiatives across sectors and industries. Firms are increasingly integrating the concept of social and environmental impact into their core business strategy. These companies recognize that ineffective use of resources, pollution, poverty, lack of education, and inequality inhibits economic growth and opportunity across the globe (Porter and Kramer 2011). To the best of our knowledge, our original contribution in this paper explains one example of interplay between business and environment using a system thinking methodology. We will illustrate positive and negative loops among business, environment, and other relevant actors that generate a win-win situation for all the actors considered in the system.

3 Methods 3.1

System Construction

Studies analyzing environment-business interaction tend to focus on how strategic environmental choices could lead to economic growth (Dowell and Muthulingam 2016). For example, converting the energy supply from fossil fuel to renewable energy can represent an advantageous investment long term. In this way of thinking, a dangerous and critical feature can be identified: the environment is considered a

66

C. A. Amadei and M. B. Borgida

mere tool to achieve a goal (i.e., increase profit). In contrast, the approach we propose reverses this viewpoint: the environment and, specifically, its physical capacity become the priority—the mean, in a typical Machiavellian view. The natural environment plays a pivotal role in the economic growth of businesses, providing resources and goods and adsorbing and transforming by-products such as pollutants and waste. Thus, the environment and its physical capacity represents the main element of the environment-business interaction and not simple tools. The world’s physical capacity can be thought of as a box where the world resources are stored (Fig. 1). The box is confined in space, as are the world resources (e.g., energy, ecosystems, materials, water, etc.), which are the main fillers of the box. Although the study could apply to any other of the world resources (i.e., box fillers), we only focus on the water dimension in order to simplify the analysis. We chose water, over other environmental dimensions, due to the several business risks associated with water and its use. For example, water mismanagement, pollution, and depletion can affect businesses via operational risk, reputational risk, or regulatory risk (Institute 2018). For instance, Hurricane Florence who recently hit the Carolinas caused serious water environmental hazard due to the overflow of hog lagoons, which contained animal waste. Moreover, consumers are becoming more and more aware of water scarcity and pollution. For instance, Nestlé faced major protests for continuing production in California while the state is in a severe water scarcity situation (Hackman 2015). In our approach, system thinking (ST) has been chosen as the main tool for analyzing the water-business interaction (WBI). ST is a robust tool, which has been developed in the last 50 years (Sterman 2000) and focuses on the interactions between constituents of a complex system. ST approach is characterized by a comprehensive view, which allows extending the boundaries of the interactions and recognizing possible other actors involved in the system. ST is also a flexible tool, which is effective in highlighting the change in the dynamics when other actors or variables are introduced in the system. In our case, we first described the status quo or the actual scenario of the WBI, and then we evaluated possible effects of improving the level of technology or education on the system. In the following section, we show and discuss how the improved scenario could lead to a more efficient use of the world resources, thus using a smaller portion of the world’s physical capacity (Fig. 1).

3.2

System Analysis

The ST representing the WBI is displayed in Fig. 2. Note that we qualitatively monitor the interactions by classifying them into positive or negative feedbacks, generating self-reinforcing (i.e., positive) or self-correcting (i.e., negative) feedback loops. Positive loops tend to amplify and enhance the output of the system; on the other hand, negative loops counteract a possible change (Sterman 2000).

Education and Technology as Levers for Sustainable Change

67

Fig. 1 Scheme illustrating the approach discussed in this study. The world’s physical capacity is analyzed with system dynamics

It is possible to notice that relationships located in the upper part of Fig. 2 represent the feedback between the water dimensions (Paredis et al. 2013). In particular, we identified three water dimensions in order to highlight the interactions of the water with other actors: (1) water use, (2) water pollution, and (3) water purification and reclaim. The first two are generally identified as the main elements when discussing water issues. On the other hand, water purification and reclamation can be identified as the mean to solve issues related to water pollution and water

68

C. A. Amadei and M. B. Borgida

Fig. 2 System thinking of the water space dynamics. Arrows describe direct feedbacks which can be positive (solid arrows) and negative (dashed arrows). Positive arrows between x and y read as: “an enhancement of x leads to an enhancement of y.” Negative arrows read as: “an enhancement of the variable x leads to a decrease of the variable y”

usage. In Fig. 2, the business part is represented by the company usage actor. It is possible to notice that the WBI is influenced by other actors such as university and research centers, nongovernmental organization (NGO), public awareness, and government and water management. The company usage is influenced by multiple actors (inlet arrows). The government can introduce regulations to limit the company water usage (Winz et al. 2009). For example, California adopted in the past a set of rules to which bars, hotels, and restaurants must oblige in order to face the drought emergency (Gambino 2015). Water management can have an impact on the company water usage by upgrading the water distribution system (Winz et al. 2009). The company usage of an environmental resources, such as water, can also be influenced by the action of NGOs. Recent examples showed the NGOs’ impact in determining standards for sustainable palm oil production (Jennifer et al. 2017). University and research centers can also contribute to varying the company water usage. For instance, new technologies requiring less amount of water and/or novel systems able to reuse wastewater can be developed by universities and then adopted, at a larger scale, by companies (Woo et al. 2016). The company usage is characterized by a single output (outlet arrow)

Education and Technology as Levers for Sustainable Change

69

determining the water use, which in turn causes water pollution and, subsequently, water purification through a causation feedback. The latter two elements (i.e., water pollution and purification) interact with the public awareness. Communities easily notice the deterioration of the wealth of the environment. The most recent example is the water crisis in Flint, Michigan. In this episode, the discovery of leadcontaminated water began with residents complaining about the water taste, smell, and appearance and various health issues (Hanna-Attisha et al. 2016). The public awareness is also positively connected to government and water management. The public can influence the regulations in terms of water management of the supply. This happened in 2011, when Italian people decided to keep their water service public via a referendum (Mattei 2013). Similarly, the public awareness and the NGO are related via positive feedbacks. NGOs can raise the public awareness in terms of water quality and water usage through their campaigns. At the same time, the participation of local communities in NGO events is crucial for both labor and active role in decision-making and project management (Davidson et al. 2007). University and research centers can also shine light on the quality of water, thus influencing the public awareness. Once more, this is what happened in Flint, where analysis from Virginia Tech reported high levels of lead in the contaminated water (Hanna-Attisha et al. 2016). It is possible to notice that that government and water management, university and research centers, NGO, and public awareness are positively connected creating positive feedback loops. Even more interestingly, the output of these feedback loops impacts negatively the water consumption and pollution and positively the water purification and reclaim. Thus, the ST elucidates how real and true collaborations between these parts could lead to a more sustainable use of water. Now we evaluate the effect of a possible boost of education and technology (regarding sustainable water use) on the status quo scenario. The improvement of existing technology can be thought of as a boost of the university and research center outputs. They play a central role in advancing the water technology and in developing engineering applications, which could be exploited by businesses. Note that an enhancement of the outputs will dictate positive feedback loops between government and water management, public awareness, and NGO, which will have as ultimate effect the reduction of water use and water pollution. Thus, it is crucial for the government to fund applied science and basic science too; the latter might look less attractive to investors in the short period, but it can lead to significant impacts in terms of innovations in the long term (Cohen et al. 2016). Correspondingly, businesses will be required to play their own part by being farseeing in adapting water sustainable technologies. In particular, a paradigm shift in businesses heavily dependent on water sources, such as farming and agriculture, is required: from productivity per unit land to productivity per unit of water (Sadler and Evans 2008). This can be favored by reducing the water losses in the distribution systems and by reusing and reclaiming the wastewater. In details, the water losses can be reduced by the advancing in automation and sensing (Wang et al. 2015), whereas advances in materials science can lead to a more efficient reuse of wastewater and overall reduction of water pollution (Werber et al. 2016). As recognized by UN, with the establishment of the UN Decade of Education for Sustainable Development (2005–2014) (UNESCO 2010), the education is a motor

70

C. A. Amadei and M. B. Borgida

for changes. We analyzed which effects an enhancement of the education in terms of sustainability will have on the use of water. Sustainability is a multidisciplinary subject, which spans from environmental studies to social sciences; this represents the largest obstacle for its establishment as a subject. Moreover, it is a subject in continuum development, characterized by high level of dynamicity. In order to overcome these challenges, sustainability cannot be taught as a set of prepackaged information but has to rely on activities that build personal awareness. These activities must stimulate a deep learning, trying to evaluate dynamics rather than fixed structures (Warburton 2003). The deep learning could be stimulated with hands-on activities or by the use of digital technology in order to engage the students in a modern fashion. In this sense, one of the most important tools is the “teaching and learning for a sustainable future” portal (UNESCO 2010), which represents a multimedia education program offered by UNESCO. Similarly, smartphone applications, which gauge students’ knowledge in terms of sustainability and help them to increase their awareness on the topic, have been developed (Weeks et al. 2016). The transition could be favored by the cooperation between NGOs and schools in order to define sustainable practices that could become part of the sustainability subject. For instance, Sustainable Jersey for Schools is an NGO providing tools and training to support schools pursuing sustainability programs (School 2018). In particular, the program offers specific activities focusing on water shortage. Businesses can also be involved in this transition; for example, Whole foods market offers learning activities for kids to teach them to have healthier lifestyle habits and consume more sustainable food (Foods 2012). At the same time, businesses could provide their workforce with the knowledge and skills necessary to make decisions in a sustainable manner. At a higher level of education, such as university, the main challenge is the establishment of sustainability as an academic field (Yarime et al. 2012). This requires a universitywide organization with the mission to connect and institutionalize the interactions among fields intersecting with sustainability. This has been achieved by several universities by creating independent research centers inside the universities (Cohen et al. 2016). For instance, Harvard University established the Center for the Environment with the mission to encourage research and education about the environment. The center offers undergraduate and graduate curricula aiming to interconnect students coming from different disciplines and exposing them to issues related to energy and environment (Harvard 2018). In summary, education is a major tool to propel a change due to its ability to positively interact with multiple actors, such as company usage, university and research center, public awareness, and NGOs.

4 Final Considerations At the time we ended this first stage of our conversation, we noticed more trends confirming the alarming trends in the field of climate change and other environmental crisis (NCA 2018); however, more data supports the positive impact created by sustainable initiatives driven by local and global firms (GRI 2018a, b).

Education and Technology as Levers for Sustainable Change

71

We invite the reader to think of further development and applications of our framework, hoping that we can encourage collaboration of professionals from multiple fields. There are emerging research agendas in several industries, so that opportunities will not be scarce (Irwin et al. 2018).

References Auth, K. (2015). The oceans: Resilience at risk. In The state of the world 2015 (pp. 79–91). Washington: Island Press. Caradonna, J. (2014). Sustainability. A history. New York: Oxford University Press. Cohen, S. A., DeFrancia, K. L., & Martinez, H. J. (2016). A positive vision of sustainability. Journal of Environmental Studies and Sciences, 6(1), 231–238. Commission, B. (1987). Towards sustainable development. In V. Hauff (Ed.), Report of the world commission on environment and development: Our common future (p. 41). Oxford: Oxford University Press. Davidson, C. H., Johnson, C., Lizarralde, G., Dikmen, N., & Sliwinski, A. (2007). Truths and myths about community participation in post-disaster housing projects. Habitat International, 31, 100–115. Dowell, G., & Muthulingam, S. (2016). Will firms go green if it pays? The impact of disruption, cost, and external factors on the adoption of environmental initiatives. Strategic Management Journal, 38(6), 1287–1304. Eccles, R. G., Ioannou, I., & Serafeim, G. (2014). The impact of corporate sustainability on organizational processes and performance. Management Science, 60(11), 2835–2857. Ernst & Young, & Boston College Center for Corporate Citizenship. (2013). Value of sustainability reporting [Online]. Accessed November 30, 2018, from https://www.ey.com/Publication/ vwLUAssets/EY_-_Value_of_sustainability_reporting/$FILE/EY-Value-of-SustainabilityReporting.pdf Foods, W. (2012). Nutrition for kids and teens [Online]. Accessed December 1, 2018, from https:// www.wholefoodsmarket.com/blog/nutrition-kids-and-teens Gambino, L. (2015). The Guardian [Online]. Accessed November 30, 2018, from https://www. theguardian.com/us-news/2015/mar/18/california-water-restrictions-restaurants-hotels-drought GRI. (2018a). Global reporting initiative [Online]. Accessed November 30, 2018, from https:// www.globalreporting.org/information/sustainability-reporting/Pages/gri-standards.aspx GRI. (2018b). Sustainable development goals [Online]. Accessed December 1, 2018, from https:// www.globalreporting.org/information/SDGs/Pages/SDGs.aspx Hackman, R. (2015). The Guardian [Online]. Accessed November 30, 2018, from https://www. theguardian.com/us-news/2015/may/20/nestle-water-bottling-california-drought Hanna-Attisha, M., LaChance, J., Sadler, R. C., & Schnepp, A. C. (2016). Elevated blood lead levels in children associated with the flint drinking water crisis: A spatial analysis of risk and public health response. American Journal of Public Health, 106, 283–290. Harvard, U. (2018). Harvard University Center for the Environment [Online]. Accessed November 30, 2018, from http://environment.harvard.edu/ Hawken, P. (2017). Drawdown. The most comprehensive plan ever proposed to reverse global warming. New York: Penguin Books. Henderson, R., Gulati, R., & Tushman, M. (2015). Leading sustainable change. Oxford: Oxford University Press. Institute, P. (2018). Business case for water sustainability [Online]. Accessed November 30, 2018, from http://pacinst.org/issues/corporate-water-stewardship/business-case-for-water-sustainability/ Irwin, E. G., et al. (2018). Bridging barriers to advance global sustainability. Nature, 1(July), 324–326.

72

C. A. Amadei and M. B. Borgida

Jennifer, L., et al. (2017). Reframing the evidence base for policy-relevance to increase impact: A case study on forest fragmentation in the oil palm sector. Journal of Applied Ecology, 54, 731–736. Knights, C. (2018). 2018 Global 100. [Online]. Accessed November 30, 2018, from https://www. corporateknights.com/reports/2018-global-100/#requestReport Kolbert, E. (2015). Fields notes from a catastrophe. London: Bloomsbury. Lovelock, J. (2009). The vanishing face of Gaia. New York: Basic Books. Matson, P., Clark, W. C., & Andersson, K. (2016). Pursuing sustainability. Princeton, NJ: Princeton. Mattei, U. (2013). Protecting the commons: Water, culture, and nature: The commons movement in the Italian struggle against neoliberal governance. South Atlantic Quarterly, 112(2), 366–376. Mullainathan, S., & Shafir, E. (2014). Scarcity: The new science of having less and how it defines our lives. New York: Picador. NCA. (2018). National climate assessment [Online]. Accessed December 1, 2018, from https:// nca2014.globalchange.gov/report Paredis, C. J. J., Bishop, C., & Bodner, D. (2013). Using system dynamics for sustainable water resources management in Singapore. Procedia Computer Science, 16, 157–166. Porter, M., & Kramer, M. K. (2011). Creating shared value. Harvard Business Review, 89(1), 62–77. Sachs, D. (2014). The age of sustainable development. New York: Columbia University Press. Sachs, J. (2015). The age of sustainable development. New York: Columbia University Press. Sadler, R. G., & Evans, J. (2008). Methods and technologies to improve efficiency of water use. Water Resources Research, 44(7), W00E04. SASB. (2018). Sustainable accounting standard board [Online]. Accessed November 30, 2018, from https://www.sasb.org/ School, S. J. F. (2018). Sustainable Jersey for school [Online]. Accessed November 30, 2018, from http://www.sustainablejerseyschools.com/ Sheffi, Y. (2018). Balancing green. When to embrace sustainability in a business (and when not to). Cambridge: MIT Press. Sloan/BCG, M. (2017). Corporate sustainability at a crossroads [Online]. Accessed November 30, 2018, from https://sloanreview.mit.edu/projects/corporate-sustainability-at-a-crossroads/ Sterman, J. (2000). Business dynamics: System thinking and modeling for a complex world. New York: McGraw-Hill. Tirole, J. (2017). Economics for the common good. Princeton: Princeton University Press. Tubiana, C. H. A. L. (2018). Earth at risk. New York: Columbia University Press. UN. (2015). Sustainable development goals. [Online]. Accessed November 30, 2018, from https:// sustainabledevelopment.un.org/?menu¼1300 UNESCO. (2010). Teaching and learning for a sustainable future [Online]. Accessed November 30, 2018, from http://www.unesco.org/education/tlsf/ Wang, Z., Song, H., Watkins, D. W., Ong, K. G., et al. (2015). Cyber-physical systems for water sustainability: challenges and opportunities. IEEE Communications Magazine, 53(5), 216–222. Warburton, K. (2003). Deep learning and education for sustainability. International Journal of Sustainability in Higher Education, 4(1), 44–56. Weeks, C., Delalonde, C., & Preist, C. (2016). Power law of engagement. In CHI conference on human factors in computing systems, Santa Clara, CA. Werber, J. R., Osuji, C. O., & Elimelech, M. (2016). Materials for next-generation desalination and water purification membranes. Nature Reviews Material, 1, 44–56. Winz, I., Brierley, G., & Trowsdale, S. (2009). The use of system dynamics simulation in water resources management. Water Resources Management, 23(7), 1301–1323. Woo, Y., et al. (2016). Effect of powdered activated carbon on integrated submerged membrane bioreactor–nanofiltration process for wastewater reclamation. Bioresource Technology, 210, 18–25. Yarime, M., et al. (2012). Establishing sustainability science in higher education institutions: Towards an integration of academic development, institutionalization, and stakeholder collaborations. Sustainability Science, 7, 101–113.

Part II

Innovation Applications

The Bezos-Gate: Exploring the Online Content of the Washington Post Katharina Weitz, Florian Johren, Lukas Seifert, Sha Li, Jiexin Zhou, Oliver Posegga, and Peter A. Gloor

Abstract After Jeff Bezos, the founder and CEO of Amazon, acquired The Washington Post in late 2013, the newspaper’s neutrality and independence from Bezos’ agenda has been in question. This paper takes a first step at exploring whether the neutrality of The Washington Post has changed after the acquisition in 2013. By analyzing the content of newspaper articles based on the sentiment and the emotional tone and by comparing results to a baseline consisting of The New York Times and The Guardian, this paper tries to establish if and how the Post’s reporting on Amazon has changed. Albeit results of this study are limited to the scope and the distortion of the analyzed data, we find an increasing number of articles about Amazon in all newspapers after the acquisition. Furthermore, it can be shown that the overall positive sentiment of The Washington Post decreases, while the emotional tone intensifies.

1 Introduction As the British economist William Brian Arthur already discovered in 1994, the economics of the tech sector is different from conventional market dynamics (Arthur 1994, 2009). Usually economic theory assumes decreasing returns of investment. The first investment has higher returns than the second, and the third one even less. The information sector, however, is based on networks, and their power increases proportional to size—or in economic terms, the marginal profit per new member increases (Wischmeyer 2018). Famous examples would be QWERT, which is the K. Weitz (*) · F. Johren · O. Posegga University of Bamberg, Bamberg, Germany e-mail: [email protected] L. Seifert · S. Li · J. Zhou University of Cologne, Cologne, Germany P. A. Gloor MIT Center for Collective Intelligence, Cambridge, MA, USA © Springer Nature Switzerland AG 2019 Y. Song et al. (eds.), Collaborative Innovation Networks, Studies on Entrepreneurship, Structural Change and Industrial Dynamics, https://doi.org/10.1007/978-3-030-17238-1_4

75

76

K. Weitz et al.

standard order of any computer keyboard. A switch to another system would involve retraining everyone used to the QWERT system and makes it too costly, even though there might be better or more efficient options already in the market. Therefore, if one asks in the Western world for the most influential companies in people’s personal life, the name of Apple, Google, Facebook, Microsoft, and Amazon will certainly show up. With the increasing digitalization and interdependency of society, the tech world replaced old power houses like oil companies, grocery stores, or car producers. Today it is almost impossible to avoid them in daily life, and any attempt of dropping those resembles near end of social life and work (Manjoo 2017). The public perception of these technology giants is often tied to the strong personalities of their founders. High-profile individuals like Mark Zuckerberg (Facebook), Bill Gates (Microsoft), Elon Musk (Tesla and SpaceX), or Jack Ma (Alibaba) are almost as famous as their brand and can be found on most Forbes rankings. Along the success of their companies, the interest in these personalities increased as well. While it is mostly the job of the tabloid press to cover the newest acquisition of an NBA team (Alibaba co-founder Joe Tsai, Paul Allen from Microsoft) or generous philanthropy project (Bill Gates), there is also a more serious aspect, namely, the impact and influence of these acquisitions. Whether it is taxation, information control, or as most recent case the influence on public opinion, governments struggle to regulate these companies. The need to limit these actions is evident, especially in the case of shaping public influence (Foroohar 2017). One event that raised strong concerns in this regard is the 2013 acquisition of a majority share of The Washington Post by Jeff Bezos, the owner of Amazon. While Bezos, who is since 2017 the richest men alive (Vinton 2017), explained his action as purely experimental independent from his main business Amazon, the sudden interest of a tech giant into a traditional print media with around half a million daily readers seemed odd. In the social media and conventional media around the globe, many questioned the neutrality of the newspaper (Bell 2013; Haughney 2013). The print media can serve as marketing tool and filter out negative press. This leads to the assumption of utility maximization where the interest of Bezos is closely linked to the further growth of his company. Five years later this project work takes a look from a big data perspective to see how much these claims were valid. The research question is: How has the coverage of The Washington Post changed with regard to Amazon after its acquisition in 2013, compared to other media? Therefore, the study represents a prototype to approach other cases in a fact-based systematic way and modern research techniques. Using common big data approaches, it can be shown that The Washington Post coverage of Amazon indeed differs significantly from other equivalent print media. The results are based on 30,846 online articles from The Washington Post, The Guardian, and The New York Times from 2008 to 2017 that are processed by different network analyzing techniques and logistic regression analyses. To distinguish between the coverage sentiment and emotionality, analyses were applied to the articles. Differences then can be compared between the various newspapers, but also of each newspaper between different time periods.

The Bezos-Gate: Exploring the Online Content of the Washington Post

77

The present paper is divided into five parts. The first section outlines the theoretical considerations. The following part will explain the methodological approach, its implementation, and in addition to that the consolidation of the data base. In the third part, the descriptive section of analysis and the logistic regression analyses are presented. In the fourth part, a summary of results and hints to future studies are given. In the fifth and final part, the conclusion, a recap about the findings, is given. The reader should follow two narratives throughout the whole paper: first, the difference between the three newspaper coverages compared to each other in each chosen time period and, second, the change in the coverage of The Washington Post after its acquisition by Jeff Bezos.

2 Background and Related Work The introductory chapter has presented the puzzle of the impartiality of media after a commercial takeover by large businesses. The focus of this paper is the specific case of The Washington Post being bought by Amazon founder Jeff Bezos in 2013. It investigates whether and how the newspaper has changed its coverage after its acquisition. In this section a review of the relevant literature shall further advance the academic discussion. This leads to the hypothesis that is formulated at the end of this section. All around the globe, mass media play a crucial role for society. They are unique due to their ability to accumulate vast amounts of resources that are concentrated on communication, the high impact on the political and economic system, as well as their reciprocal character (Jamieson and Campbell 1992). Multiple studies have confirmed the influence of mass media on the civil society by informing the public and shaping the debate of society (McCombs and Shaw 1972; McQuail 1977). The public influence appears to be particularly strong with regard to politics and shopping behavior. Chiang and Knight (2011) showed that biased newspapers have an effect on the voting behavior of their readership. King et al. (2017) even identify an influence not just in times of elections but throughout the whole political cycle. While a lot is known about the advertisement power of the classical mass media like television on consumer attitudes (Priya et al. 2010), the Internet raised the magnitude on another level (Boulianne 2015). The time spent on the Internet has worldwide increased; in 2008 75% of the Internet users use online social media every day (Kaplan and Haenlein 2010). Therefore, firms invest a greater share of their budget to reach out on these platforms through advertisement and research how to improve consumer engagement (Lee et al. 2018). State authorities are aware of this relation. The Federal Communication Commission of the United States aims to diversify the media landscape and diffuse ownership among media firms to enable the public to have access to different perspectives (Gentzkow and Shapiro 2010). This brief review of relevant literature illustrates the significance of today’s media/business relations. While the approach of the present project is explorative

78

K. Weitz et al.

by nature, it can definitely be placed within the current academic and public debate. The influence by mass media is channeled through two pathways: 1. Directly, through the content of an article. This may include writing about current affairs of a company or blaming competitors, but also the absence of negative reports about pressing issues. 2. Indirect, through the advertisement placed around the actual content/article. It is assumed that The Washington Post has a motive to change its coverage toward the interest of its new owner. This may happen by a better presentation of Amazon in the media landscape and therefore directly or indirectly pursue influence on their readers. This paper focusses on the first pathway. In conclusion the hypothesis is derived: It is more likely that The Washington Post (online) presents Amazon in a more positive way than other newspaper.

The hypothesis will test the strength of the owner and media relationship. Since we assume a strong tie, we should observe a significant difference in the online content. The Washington Post is assumed to frame stories more positively with regard to Amazon. The hypothesis is not valid if the published content shows no difference among the newspapers.

3 Methods Our approach to answer the research question is by explorative quantitative methods. The work process was divided into three steps, namely, data aggregation, data preparation, and data analysis. As the focus of analysis is on comparing the content of news articles of different news companies, the first step is to gather the content of the news articles. Since the influence of Jeff Bezos on articles from The Washington Post was investigated, all news articles from The Washington Post mentioning the word Amazon were collected. Additionally, news articles about Amazon from The Guardian and The New York Times were collected to build a general baseline that can be used to compare The Washington Post against. As a second step, the dataset was prepared, and variables needed for the analysis were calculated. The data were cleaned up, and the key values for each article such as the article’s sentiment, the article’s emotions, and the number of times the word Amazon is mentioned within an article were calculated. In the last step of the approach, values calculated in step 2 were used to analyze the dataset and perform logistic regression. In order to aggregate the different news articles, a Python-based script was used to crawl through news articles on washingtonpost.com, theguardian.com, and nytimes.com and scrape the article’s information. Article’s information included the title, the content, the news company, and the publishing date (see Table 1). All three of the news companies provide an archive that enables it to filter articles based on the keyword “Amazon.” After the data aggregation, the raw dataset includes 5946 articles of The Washington Post, 17,191 articles of The Guardian, and 14,027 of The New York Times.

The Bezos-Gate: Exploring the Online Content of the Washington Post

79

Table 1 Structure of the raw dataset ID ... 42 ...

Publishing date ... 2017-1109 ...

Company ... Washington Post ...

Title ... Kindle with your kale? Amazon to open shops in Whole Foods ...

Content ... You may see more of Amazon. . .. ...

By cleaning the dataset, the news articles were reduced to 5484 articles of The Washington Post, 17,145 articles of The Guardian, and 8217 articles of The New York Times. The huge reduction of articles of The New York Times is due to the fact that the archive of nytimes.com included deprecated links to the detail page of an article. Since the timespan of publications varies based on the news company, a further compress of the dataset was made. To make the news articles more comparable, the time frame of the investigation was limited. In the analysis part, only articles with a publishing date between 2008 and late 2017 were used. To further prepare the dataset for the analysis, the IBM Watson® sentiment analysis tool was used to compute the overall sentiment, the target sentiment, and the emotional tone of each news article (see Table 2). The overall sentiment depicts the sentiment of the whole article, whereas the target sentiment specifically focuses on the sentiment of the target keyword “Amazon.” Both the target sentiment and the overall sentiment are derived as values between
1 and 1, where
1 describes a very negative article and 1 describes a very positive article. The emotional tone is depicted by five different emotions following Ekman’s list of basic emotions (Ekman et al. 1972), namely, fear, joy, sadness, anger, and disgust. The emotion surprise was not supported by IBM Watson®. Each emotional value ranges between 0 and 1, where 0 depicts the absence of the emotion and 1 depicts the absolute presence of an emotion within the article.

4 Results We started by analyzing the frequency of publications and creating word clouds of prominent occurrences on the frequency chart. Since the word clouds as well as the frequency did not yield any exceptional results, the focus was shifted to sentiment analysis and to an analysis of the emotions. Finally, we conducted logistic regression. Every step of analysis is depicted in detail below.

4.1

Word Clouds

As a first step, word clouds were generated for The Washington Post articles containing the keyword “Amazon.” As depicted in Fig. 1, the word cloud on the

Publishing date ... 2017-1109

...

ID ... 42

...

...

Company ... Washington Post

Title ... Kindle with your kale? Amazon to open shops in Whole Foods ...

Content ... You may see more of Amazon. . .. ... ...

Overall sentiment ... 0.9110

Table 2 Structure of the cleaned dataset. Includes calculations for sentiment and for emotions

...

Target sentiment ... 0.8897

...

Fear ... 0.00845

...

Joy ... 0.6431

...

Sadness ... 0.0823

...

Anger ... 0.1877

...

Disgust ... 0.0192

80 K. Weitz et al.

The Bezos-Gate: Exploring the Online Content of the Washington Post

81

Fig. 1 Word clouds of The Washington Post articles before (left) and after acquisition (right)

left shows the words used before the acquisition, and the word cloud on the right shows the used words after the acquisition. The size of the words has no meaning. Even though some of the words within the word cloud have changed, the word occurrence seems not to be a good indicator to show the influence of Jeff Bezos on The Washington Post. A different frequency in the wording is most probably due to changes in the overall news coverage.

4.2

Frequency of Publications

To get a first overview, the frequency at which each newspaper has published news articles about Amazon was analyzed. Before visualizing the frequency, it is assumed that the influence of Jeff Bezos on The Washington Post would be shown in an unusual behavior on the chart. It is assumed that the frequency either should have dropped or should have increased after the announcement of Jeff Bezos buying The Washington Post in 2013. But as depicted in Fig. 2, none of the expected behavior occurs. The graph shows the accumulated publications of news articles about Amazon of The Washington Post, The New York Times, and The Guardian. While the publications of The New York Times and The Guardian increase, the publications of The Washington Post start to fluctuate. Even though it seems like the frequency of publications of The Washington Post increases after 2014, the frequency of publication seems to be a poor measurement value. Since end of 2015 the number of publications increases drastically, it is assumed that the dataset or the data source might be influenced by some collection error. For instance, it could be that the archives of the news companies include more links to recent news articles and only news articles with a certain degree of popularity are stored in the archive on a long-time basis.

4.3

Sentiment Analysis

For the sentiment analysis, using IBM Watson® sentiment analyzer, only news articles that contain the keyword “Amazon” within their headline were included,

82

K. Weitz et al. 4000

Washington Post New York Times

No. of Publications about Amazon

The Guardian 3000

2000

1000

0 2008

2010

2012

2014

2016

Time

Fig. 2 Frequency of articles about Amazon in the three different newspapers

to minimize distortion. This ensures that the sentiment analysis only includes news articles that have Amazon as their main topic, yielding more accurate results. First the values of the target sentiment for each newspaper were plotted, and a polynomial function was fit to the dataset (see Fig. 3). By using the fit function, the best matching curve to the data points is created, and it returns a coefficient that minimizes the squared error. Figure 3 depicts the polynomial functions of each news company. It shows that on average The Guardian writes most consistently in a positive manner about Amazon. The New York Times is also consistent in reporting, but talks the most negative about Amazon. The sentiment of The Washington Post fluctuates the most, talking very positively about Amazon between 2012 and 2014 but also talking more negatively about Amazon between 2015 and 2017. It is important to mention that the positive peak in the sentiment of The Washington Post occurs exactly around the time frame when Jeff Bezos announced the acquisition of The Washington Post. Nevertheless, those findings have to be considered with a grain of salt, since (a), (b), and (c) in Fig. 4 emphasize that all data functions show a very high standard deviation.

4.4

Emotion Analysis

To further distinguish the differences between the newspapers, emotions such as joy, sadness, fear, anger, and disgust were used to calculate the differences of each emotion before and after the acquisition. As depicted in Fig. 5 for The New York Times and The Guardian, the emotions disgust, fear, and sadness have been reduced

The Bezos-Gate: Exploring the Online Content of the Washington Post

0.8

83

Washington Post New York Times The Guardian

0.6

Sentiment

0.4

0.2

0.0

–0.2

2010

2011

2012

2013

2014 Time

2015

2016

2017

2018

Fig. 3 Average sentiment values over time for the three newspapers: The Washington Post, The New York Times, and The Guardian

after The Washington Post was acquired by Jeff Bezos, while for The Washington Post all emotions except the sadness increased. Since also the value joy increased by as much as almost 6%, it seems like The Washington Post writes more emotional when talking about Amazon compared to the other two news companies. Furthermore, those findings fit well with the results of the sentiment analysis, because especially negative emotions like anger, disgust, and fear increased, while the target sentiment decreased. To take a closer look to the changing emotions in Amazon-related articles, IBM SPSS® was used to do statistical analysis. To check whether The Guardian and The New York Times differed in reporting on Amazon before or after the takeover of Jeff Bezos from The Washington Post, we chose logistic regression as the evaluation method. In the logistic regression, the type of newspaper served as a dependent variable; the emotions joy, fear, sadness, anger, and disgust served as independent variables: Newspaper (0/1) ¼ w1  sadness + w2  fear + w3  disgust + w4  joy. The prerequisites for logistic regression were fulfilled for the calculated models. The predictors were centered on the mean, and the values were z-standardized. The Hosmer-Lemeshow test showed significant results for the “after acquisition” models, which points to a poor model fitting. Due to the high number of cases, the results are nevertheless shown. The difference in emotions between The Washington Post (coded as 1) and The New York Times and The Guardian (both coded as 0) was compared before and after the acquisition by Jeff Bezos. Therefore, two kinds of

84

K. Weitz et al.

a 1.5

Sentiment

1.0

0.5

0.0

–0.5

–1.0 2010

2011

2012

2013

2014

2015

2016

2017

2018

Time

b 0.4

Sentiment

0.2

0.0

–0.2

–0.4

–0.6 2010

2011

2012

2013

2014

2015

2016

2015

2016

2017

2018

Time

c

1.25 1.00

Sentiment

0.75 0.50 0.25 0.00 –0.25 –0.50 –0.75 2010

2011

2012

2013

2014

2017

2018

Time

Fig. 4 Sentiment values over time for the three newspapers: (a) The Washington Post, (b) The New York Times, (c) and The Guardian. Intervals represent the standard deviation

The Bezos-Gate: Exploring the Online Content of the Washington Post 0.06

85 Washington Post New York Times The Guardian

Difference in Emotions

0.04

0.02

0

–0.02

–0.04

anger

disgust

fear

joy

sadness

Fig. 5 Difference in emotions in Amazon-related articles after the acquisition, compared between the three newspapers: The Washington Post, The New York Times, and The Guardian

Table 3 Description of the dataset Time Before acquisition After acquisition

Total sample (N ) 941 1644

Used sample (n) 911 1569

Missings 30 75

models were constructed: one model before the acquisition by Jeff Bezos in the year 2013 and one model after the acquisition (see Table 3). All five emotion parameters (joy, fear, sadness, disgust, and anger) were included in the calculation. The logistic regression shows significant results before the acquisition for all emotions, except disgust (see Table 4). The regression model shows overall significance, χ 2 (5) ¼ 136.136, p < .001, and a respective explanatory power model of 19.2% (Nagelkerke R2), which means 19.2% of the variance in newspaper differences in reporting emotionally about Amazon before the acquisition by Jeff Bezos. The model correctly classified 72.1% of cases. The first investigation for the data after acquisition shows that only sadness and anger remain significant (see Table 5). The regression model was statistically significant, χ 2 (5) ¼ 15.807, p < .05. The model explained 0.19% (Nagelkerke R2) of the variance in newspaper differences in reporting emotionally about Amazon after the acquisition by Jeff Bezos. The model correctly classified 88.3% of cases. Since there were no previous assumptions that all emotions are relevant as predictors for the model, the next step was to construct two stepwise models, one for the articles before the acquisition and one for the articles after the acquisition by Jeff

86

K. Weitz et al.

Table 4 Before acquisition—logistic regression with all emotions as predictors Sadness Fear Disgust Joy Anger

B
0.373
0.450
0.132
0.440
0.554

SE 0.077 0.105 0.094 0.077 0.116

Wald’s χ 2 23.558 18.274 1.961 32.962 22.987

df 1 1 1 1 1

p .000* .000* .161 .000* .000*

Exp(B) 0.689 0.637 0.876 0.644 0.574

Nagelkerke R2 0.192

*p < .001 Table 5 After acquisition—logistic regression with all emotions as predictors Sadness Fear Disgust Joy Anger

B
0.187
0.057 0.010
0.071
0.285

SE 0.081 0.097 0.094 0.087 0.110

Wald’s χ 2 5.321 0.344 0.011 0.670 6.752

df 1 1 1 1 1

p .021* .558 .917 .413 .009**

Exp(B) 0.829 0.944 1.010 0.932 0.752

Nagelkerke R2 0.019

*p < .05, **p < .01 Table 6 Before acquisition—logistic regression after step-by-step model construction Sadness Fear Joy Anger

B
0.383
0.472
0.430
0.592

SE 0.076 0.105 0.076 0.114

Wald’s χ 2 25.131 20.138 31.941 27.023

df 1 1 1 1

p .000* .000* .000* .000*

Exp(B) 0.682 0.624 0.650 0.553

Nagelkerke R2 0.190

*p < .001 Table 7 After acquisition—logistic regression after step-by-step model construction Sadness Anger

B
0.180
0.270

SE 0.077 0.103

Wald’s χ 2 5.387 6.855

df 1 1

p .020* .009**

Exp(B) 0.835 0.763

Nagelkerke R2 0.018

*p < .05, **p < .01

Bezos. Here, the logistic regressions show significant results for the remaining emotions before and after the acquisition. Before the acquisition, the regression model was statistically significant, χ 2 (1) ¼ 26.720, p < .001. The model explained 19.0% (Nagelkerke R2) of the variance in newspaper differences in reporting emotionally about Amazon before the acquisition by Jeff Bezos (see Table 6). The model correctly classified 66.2% of cases. After the acquisition, the regression model was statistically significant, χ 2 (1) ¼ 5.256, p < .01. The model explained 0.18% (Nagelkerke R2) of the variance in newspaper differences in reporting emotionally about Amazon after the acquisition by Jeff Bezos (see Table 7). The model correctly classified 88.3% of cases.

The Bezos-Gate: Exploring the Online Content of the Washington Post

87

5 Discussion The analysis of the data should answer the hypothesis presented at the beginning. The hypothesis was: It is more likely that The Washington Post presents Amazon in its online article in a more positive way than other newspaper.

A general finding that The Washington Post reports more positive in news articles about Amazon after the takeover by Jeff Bezos could not be found at first glance. While the overall sentiment has decreased, the emotional degree has increased. It seems that after the takeover by Jeff Bezos, The Washington Post reports more negatively about Amazon but in a more emotional manner than other newspapers. Furthermore, comparing the results of the logistic regression before and after the acquisition raises the assumption that the decreasing sentiment might be due to The Washington Post getting closer to The New York Times and The Guardian.

5.1

Limitations

Due to its complexity, big data analysis offers many possibilities, but also the danger of misinterpretation of the data (Butts 2009). Therefore, it is emphasized that the results of this paper should only be seen as the beginning of a comprehensive analysis. Even though the dataset has been cleaned and prepared for analysis purposes, more extensive filtering is required to yield better results and minimize distortion. All filtering of this paper has been done by only applying queries on the database. Results could improve or even completely change by manually reviewing every news article and adjusting the selection based on human judgment. The results of the logistic regression show that even before Jeff Bezos’ acquisition of The Washington Post, significant differences between the newspapers need to be examined and analyzed in more detail. It is obvious that findings of this project are limited due to the fact that newspapers naturally differ. Already Armstrong (2004) noted that articles in newspapers are influenced by the reporters and their personal attitudes. In addition, newspaper-specific differences can also be found in the presentation and reporting of events related to the political orientation of the newspaper (Feldman et al. 2017). The project work shows an initial trend that reporting about Amazon has increased following the takeover by Jeff Bezos in The Washington Post. However, it remains unclear exactly what the causes are. One simple explanation could be the overall expansion of online content measured in amount of articles per year.1 The newspaper market is experiencing a vast transformation from print to online. A higher number of Amazon articles could correlate with the growth of articles in 1

https://wordpress.com/activity/posting/

88

K. Weitz et al.

general. This quasi-hypothesis, however, requires more precise data on the online publications of the newspapers investigated. Additionally, this initial analysis does not provide any information as to what the increase in articles about Amazon means in terms of advertisement content, for example, what type of advertising is concerned and which products are advertised. For that, a clear definition of what is meant by advertising is needed. Such a clear definition is necessary in order to look specifically for advertising within the articles. Such “hidden” advertisements could make it possible to advertise products that can be of interest to the reader of the article and can be categorized as “Ad Type” described by Rodgers and Thorson (2000), where the authors state it as an indicator of possible consumer reactions. In addition, to analyze the text content of the newspaper articles for advertisements, banners or advertisements displayed on the online pages of the newspaper should be taken into account. Furthermore, the analysis of advertising should take place not only at The Washington Post. A detailed analysis of Amazon advertising in the other newspapers need to be done, too. Only then a general trend toward more advertising can be ruled out. Factors such as the attitudes of the article authors also remain hidden. Armstrong (2004) shows that newspaper authors have a clear influence on the content of the articles they have written. Another aspect that is not considered in this work but is very important is the reaction of users to articles. In theoretical models of advertisements like the Integrating Advertising Model (Rodgers and Thorson 2000), not only the advertiser but also the consumer is included. It is assumed that the possibilities of the online presentation on the Internet allow an interaction between consumer (reader) and advertiser (online newspaper), which again have influence on the authors and the future articles of the newspaper. The social impact on future behavior has already been demonstrated on social media platforms (Berry and Taylor 2017). Complex models should therefore be used for in-depth analysis, which in addition to the diversity of the newspapers also model the influence and reactions of the readers.

5.2

Future Research

The results of this project work open up exciting future research questions and projects. For example, a detailed Amazon advertising analysis in other newspapers could be done. Possible questions could be: Did the purchase of The Washington Post by Jeff Bezos influence the advertisements in this newspaper? Has the acquisition influenced Amazon advertisements in other newspapers? How and how often does Amazon advertise there? This study looked at the coverage of only three different newspapers. One could broaden this perspective by either including more national newspapers to the dataset or by extending the scope to an international perspective and analyzing newspapers from other (non-English language) countries. It has already been shown that readers of newspapers differ from country to country (Elvestad and Blekesaune 2008). Therefore, it could be assumed that media coverage differs from country to country, too.

The Bezos-Gate: Exploring the Online Content of the Washington Post

89

The influence of readers on readers must also be taken into account through the commentary functions on the online pages of the “consumer-controlled” side, as each of the newspapers considered in this paper offers the reader the opportunity to communicate with each other and with the authors using the comment function. It also offers advertisers the opportunity to receive active feedback from consumers. For social media platforms, such an influence could already be demonstrated by interaction (Mangold and Faulds 2009), and for other topics, social influence on newspapers could be shown, too (Golan 2006). Furthermore, it could be investigated, if there is a bias between newspapers when they are being taken over by other media like Amazon.

6 Conclusion In this paper it could be shown that the takeover of The Washington Post by the managing director of Amazon, Jeff Bezos, not only generated a social debate about the linking of reporting with advertising but also shows an increase in articles about Amazon. Reasons for this do not necessarily have to be the takeover but can also be due to a general increase of online content. This paper lays an important foundation for further research questions, which should focus primarily on the placement of advertising. Furthermore, it is important to include the influence and reaction of users in further studies.

References Armstrong, C. L. (2004). The influence of reporter gender on source selection in newspaper stories. Journalism & Mass Communication Quarterly, 81(1), 139–154. https://doi.org/10.1177/ 107769900408100110 Arthur, W. B. (1994). Increasing returns and path dependence in the economy. Ann Arbor: University of Michigan Press. Arthur, W. B. (2009). The nature of technology. What it is and how it evolves. Simon and Schuster. Bell, E. (2013). Jeff Bezos and the Washington Post: A marriage of old media and new money. Retrieved February 13, 2018, from https://www.theguardian.com/commentisfree/2013/aug/06/ jeff-bezos-washington-post-media-marriage Berry, G., & Taylor, S. J. (2017). Discussion quality diffuses in the digital public square. In Proceedings of the 26th International Conference on World Wide Web: 1371–1380. https:// doi.org/10.1145/3038912.3052666 Boulianne, S. (2015). Social media use and participation. A meta-analysis of current research. Information, Communication & Society, 18(5), 524–538. https://doi.org/10.1080/1369118X. 2015.1008542 Butts, C. T. (2009). Revisiting the foundations of network analysis. Science, 325(5939), 414–416. https://doi.org/10.1126/science.1171022 Chiang, C.-F., & Knight, B. (2011). Media bias and influence. Evidence from newspaper endorsements. The Review of Economic Studies, 78(3), 795–820. https://doi.org/10.1093/restud/rdq037

90

K. Weitz et al.

Ekman, P., Friesen, W. V., & Ellsworth, P. (1972). Emotion in the human face. Guidelines for research and an integration of findings. New York: Pergamon Press. Elvestad, E., & Blekesaune, A. (2008). Newspaper readers in Europe. A multilevel study of individual and national differences. European Journal of Communication, 23(4), 425–447. https://doi.org/10.1177/0267323108096993 Feldman, L., Hart, P. S., & Milosevic, T. (2017). Polarizing news? Representations of threat and efficacy in leading US newspapers’ coverage of climate change. Public Understanding of Science, 26(4), 481–497. https://doi.org/10.1177/0963662515595348 Foroohar, R. (2017). Why we need to regulate the tech platforms. Retrieved May 10, 2018, from https://www.ft.com/content/84f402ac-bfc0-11e7-b8a3-38a6e068f464 Gentzkow, M., & Shapiro, J. M. (2010). What drives media slant? Evidence from U.S. daily newspapers. Econometrica, 78(1), 35–71. https://doi.org/10.3982/ECTA7195 Golan, G. (2006). Inter-media agenda setting and global news coverage. Assessing the influence of the New York Times on three network television evening news programs. Journalism Studies, 7 (2), 323–333. https://doi.org/10.1080/14616700500533643 Haughney, C. (2013). Bezos, Amazon’s founder, to buy. The Washington Post. Retrieved February 13, 2018, from http://www.nytimes.com/2013/08/06/business/media/amazoncom-founder-tobuy-the-washington-post.html Jamieson, K. H., & Campbell, K. K. (1992). The interplay of influence. News, advertising, politics, and the mass media (3rd ed.). Belmont, CA: Wadsworth Publishing. Kaplan, A. M., & Haenlein, M. (2010). Users of the world, unite! The challenges and opportunities of Social Media. Business Horizons, 53(1), 59–68. https://doi.org/10.1016/j.bushor.2009.09.003 King, G., Schneer, B., & White, A. (2017). How the news media activate public expression and influence national agendas. Science, 358(6364), 776–780. https://doi.org/10.1126/science. aao1100 Lee, D., Hosanagar, K., & Nair, H. S. (2018). Advertising content and consumer engagement on social media. Evidence from Facebook. Management Science, 64(11), 1–17. https://doi.org/10. 1287/mnsc.2017.2902 Mangold, W. G., & Faulds, D. J. (2009). Social media. The new hybrid element of the promotion mix. Business Horizons, 52(4), 357–365. https://doi.org/10.1016/j.bushor.2009.03.002 Manjoo, F. (2017). Tech’s frightful five: they’ve got us. Retrieved February 10, 2018, from https:// www.nytimes.com/2017/05/10/technology/techs-frightful-five-theyve-got-us.html McCombs, M. E., & Shaw, D. L. (1972). The agenda-setting function of mass media. Public Opinion Quarterly, 36(2), 176–187. https://doi.org/10.1086/267990 McQuail, D. (1977). The influence and effects of mass media. In Mass communication and society (pp. 70–94). London: Edward Arnold. Priya, P., Kanti Baisya, R., & Sharma, S. (2010). Television advertisements and children’s buying behaviour. Marketing Intelligence & Planning, 28(2), 151–169. https://doi.org/10.1108/ 02634501011029664 Rodgers, S., & Thorson, E. (2000). The interactive advertising model. Journal of Interactive Advertising, 1(1), 41–60. https://doi.org/10.1080/15252019.2000.10722043 Vinton, K. (2017). Amazon CEO Jeff Bezos is the richest person in the world – again. Retrieved February 13, 2018, from https://www.forbes.com/sites/katevinton/2017/10/27/amazon-ceo-jeffbezos-is-the-richest-person-in-the-world-again/#7d4ed18e1948 Wischmeyer, N. (2018). Die Geister, die er rief. Retrieved February 11, 2018, from https://www. brandeins.de/magazine/brand-eins-wirtschaftsmagazin/2018/reset/die-geister-die-er-rief

Identifying Tribes on Twitter Through Shared Context Peter A. Gloor, Andrea Fronzetti Colladon, Joao Marcos de Oliveira, Paola Rovelli, Manuel Galbier, and Manfred Vogel

Abstract This paper introduces Tribefinder, a novel system able to reveal Twitter users’ tribal affiliations. Tribefinder establishes to which tribes individuals belong through the analysis of their tweets and the comparison of their vocabulary. These tribal vocabularies are previously generated based on the vocabulary of tribal influencers and leaders selected using Tribecreator. To demonstrate its functionality, in the case presented in this paper, the system was calibrated in three specific tribal macro-categories: alternative reality, lifestyle, and recreation. Apart from describing the methodology we used to create this system, we also provide some practical examples of its use, thus giving a first indication of its potential. Finally, we present the results of the adoption of a t-SNE visualization approach, useful to verify whether tribe members cluster closely together.

P. A. Gloor MIT Center for Collective Intelligence, Cambridge, MA, USA e-mail: [email protected] A. Fronzetti Colladon Department of Engineering, University of Perugia, Perugia, Italy e-mail: [email protected] J. M. de Oliveira galaxyadvisors AG, Aarau, Switzerland e-mail: [email protected] P. Rovelli (*) Faculty of Economics and Management, Free University of Bozen-Bolzano, Bolzano, Italy e-mail: [email protected] M. Galbier · M. Vogel University of Applied Sciences Northwestern Switzerland, Brugg, Switzerland e-mail: [email protected]; [email protected] © Springer Nature Switzerland AG 2019 Y. Song et al. (eds.), Collaborative Innovation Networks, Studies on Entrepreneurship, Structural Change and Industrial Dynamics, https://doi.org/10.1007/978-3-030-17238-1_5

91

92

P. A. Gloor et al.

1 Introduction A tribe is “a network of heterogeneous persons linked by a shared passion or emotion” (Cova and Cova 2002). In other words, a tribe is a means whereby individuals experience a sense of community and share strong emotional links, common culture, passions, and vision of life (Cova 1996; Cova and Cova 2002; Richardson 2013). Individuals break up in several different tribes, and each of them may belong to many smaller and larger tribes, playing different roles and wearing different masks (Cova 1996; Cova and Cova 2002). Individuals’ expressed behaviors reveal to what tribes they belong and how they perceive their own identity (e.g., Garry et al. 2008). Indeed, each tribe has its own peculiarities, behaviors, rituals, traditions, myths, values, beliefs, hierarchy, and vocabulary (Cova and Pace 2006), which support the identification of individuals’ tribal affiliations. Observing the emergence of “tribalism” (Bauman 1990; Maffesoli 1996), it became clear that understanding its “tribes” is essential for firms’ survival (e.g., Holzweber et al. 2015), being especially important for marketing (e.g., Goulding et al. 2013; Kozinets 1999). To extend traditional marketing strategies (Addis and Podesta 2005; Canniford 2011), scholars have started suggesting to firms to rethink their marketing activities (Cova and Cova 2002; Moutinho et al. 2007), taking in account the existence and behavior of their consumer tribes—i.e., “tribal marketing” (Cova and Cova 2002). Tribes’ characteristics may indeed affect the success of a marketing campaign, even if few studies exist so far on how they can be used as a strategic resource (Cova and Cova 2002). Marketing actions should be designed depending on the tribes that have to be addressed given the characteristics of the firm, its brand, and the product or service it offers (Moutinho et al. 2007). At the same time, knowing what types of tribes are particularly attracted by a specific product or brand may be a powerful instrument to improve marketing of this product or brand. In doing so, firms have the possibility to design their marketing actions in line with the individual and social needs of tribes’ members (e.g., Cova 1996; Holzweber et al. 2015), thus maximizing the probability of success. However, the identification of tribes is difficult and requires different and special efforts (Cova and Cova 2002). Moreover, the advent of the Internet and the growing use of social media as marketing instruments (Burton and Soboleva 2011) challenge even more the identification of the so-called virtual tribes, meaning tribes that nowadays form by communication technologies (Cova and Pace 2006). This, in turn, calls for new methodologies to properly identify these virtual tribes. This is particularly true given the limits of the traditional approaches used by existing studies on consumer tribes—e.g., ethnography and nethnography (Cova and White 2010; Goulding et al. 2013; Hamilton and Hewer 2010), focus groups (Dionísio et al. 2008; Moutinho et al. 2007), interviews (Cova and Cova 2002; Cova and Pace 2006; Holzweber et al. 2015), and surveys (Taute and Sierra 2014)—which do not allow to automatically and systematically identify virtual tribes and their characteristics. Our paper presents a novel system, called Tribefinder, to identify virtual tribes (hereafter: tribes). Leveraging Twitter, it analyzes an individual’s tweets and

Identifying Tribes on Twitter Through Shared Context

93

categorizes her/him into tribes belonging to three specific tribal macro-categories: alternative reality, lifestyle, and recreation (De Oliveira and Gloor 2018). While these macro-categories have been chosen just as examples to demonstrate how the system works, Tribefinder can be easily extended to other macro-categories depending on the user’s needs. Through this system it is possible to automatically classify any individual into her/his tribal affiliations by any macro-category that is of interest for the analyst. After presenting our novel system and its validation, we provide a t-SNE visualization of tribe members. t-SNE offers an intuition on the quality of the tribe classification generated by Tribefinder by verifying whether individual tribe members cluster closely together.

2 Developing a System to Reveal Tribes The continuous stream of tweets is an important source of information (e.g., Bringay et al. 2011), which offers a powerful setting for studying and identifying tribes of individuals. The goal of the proposed system Tribefinder is to categorize Twitter users into alternative orthogonal tribes. This is made possible by extracting information about key people, brands, and topics from their tweets. Tribefinder provides as output the tribal affiliations of an individual, consistent with three tribal macrocategories: alternative reality, lifestyle, and recreation. These three specific macrocategories have been chosen to provide an example of how the system functions. This is not intended to be a limitation of the system. Tribefinder can be personalized depending on the interests of who uses the system. Applying the same methodology, Tribefinder is currently being extended to similarly identify user-defined tribal macro-categories. The Tribefinder system consists of two main components (Fig. 1): the tribe creation and the tribe allocation modules. To create and then train Tribefinder, a user first has to identify key individuals who represent the different predefined tribes

Fig. 1 Tribefinder system architecture

94

P. A. Gloor et al.

for each tribal macro-category (e.g., the tribes nerd, fatherlander, spiritualist, and tree hugger for the macro-category alternative reality). Through this process, a large sample of Twitter users is generated belonging to each of these newly created tribes, defined by the concepts, ideas, and artifacts that may describe them. A tribe can be idealized as a concept, idea, or artifact that its members believe in or like (De Oliveira and Gloor 2018). More specifically, this search is performed using Tribecreator (De Oliveira and Gloor 2018), a Web tool that allows users to automatically find individuals by keywords expressing concepts, ideas, and beliefs, using four search functions. New tribe users can be searched based on the match between the tribe’s general characteristics and the individual’s (1) Twitter profile description, (2) tweets, (3) followers, and (4) friends (i.e., those whom s/he follows).1 The Twitter timeline of the users that likely belong to each predefined tribe is then gathered by Tribecreator. This collected data is subsequently used to create a tribal vocabulary and the machine-learning model to find the tribal affiliations of a given individual. However, our system also utilizes this information to get a preliminary understanding of the tribal affiliations of the individuals previously extracted. The characteristics of such a newly created tribe can be visualized in three ways. First, Tribecreator draws a network of the tribe’s members, to have a first idea about the most influential individuals. Second, a hashtag word cloud can be generated, to identify the top hashtags. Third, the most popular posted links can be shown. As mentioned before, once a tribe has been created, its tribal vocabulary is computed. This final step to make the system learn on how to associate random individuals with specific tribes consists of the analysis of the language these influential tribal leaders use through deep learning. In so doing, classifiers are created using embedding and LSTM (long short-term memory) models. Specifically, these classifiers work by collecting the Twitter feeds of all the users from the tribes that Tribefinder is training on. On these, embedding is applied to map words into vectors, which are then used as input for the following LSTM models. LSTM models are deep learning models specially designed to analyze sequential data, which are used in this case to analyze not only what individuals say on social media but also how they say it. The model thus tries to learn how to predict a tribal affiliation for a single tweet. Once a tribe is predicted for each tweet, Tribefinder sums up the result to have a tribe distribution for the user timeline. In other words, analyzing recurring concepts in the tweets of influential leaders, Tribefinder identifies the textual patterns that characterize each tribe and generates a specific tribal vocabulary. Table 1 summarizes tribal macro-categories and actual tribes we identified. Specifically, Tribefinder uses three macro-categories to define individuals’ tribal affiliations (i.e., alternative reality, lifestyle, and recreation). Looking, for instance, at the alternative reality to which individuals belong, Tribefinder separates them into four tribes: nerds, tree huggers, spiritualists, and fatherlanders. The so-called nerds are technocrats who believe in a global world ruled by capital and technology; the

1 For followers and friends, their tweets are analyzed to understand whether the individual is connected with accounts that post tweets on topics related to the tribe.

Identifying Tribes on Twitter Through Shared Context

95

Table 1 Tribefinder tribal macro-categories and tribes Tribal macrocategory Alternative reality

Tribes Fatherlander

Nerd

Spiritualist

Tree hugger

Lifestyle

Fitness

Sedentary Vegan Yolo

Recreation

Art

Fashion Sport Travel

Description They believe in God and fatherland and that their fatherland is the best one. They cling to the good old times, hold the idea of the family in high regard, and have little time for foreigners They believe that progress, science, and technology are a blessing. They want to overcome death and colonize Mars. They are fans of globalization and network with each other They believe in a subjective experience of a sacred dimension. They find strength in contemplation, and their behavior is driven by the search for sacred meaning They believe in the limits of growth and in the protection of nature. They challenge some elements of technological progress (e.g., gene manipulation) and welcome others (e.g., alternative energies) They love doing sports and are addicted to training. They show an almost compulsive engagement in any form of physical exercise Opposite to the fitness addicted, they are characterized by much sitting and little physical exercise They follow a plant-based diet avoiding all animal foods, as well as avoiding using animal products They follow the motto “You only live once,” and they think that one should make the most of the present without worrying about the future (“carpe diem”). As a consequence, they often adopt impulsive and reckless behavior They are interested in any form of art (e.g., paintings, sculptures, music, dance, literature, films), of which they appreciate the beauty and emotional power They are interested in popular or the latest style of clothing, hair, decoration, or behavior They love watching any kind of sport on TV and attending sports events. Some also actually like to practice these sports They love travelling around in the world, for both pleasure and business, experiencing different cultures and environments

tree huggers fight for protecting the environment, while the spiritualists are individuals who mainly focus their attention on the spiritual side of things. On the opposite side, the fatherlanders are ultra-patriots who want to recreate the national states of the early twentieth century. Using Tribefinder and the tribal vocabulary it learned, it is now possible to establish the tribal affiliations of every Twitter user. In practice, Tribefinder analyzes the individual’s word usage in her/his tweets and then assigns the corresponding alternative reality, lifestyle, and recreation tribal affiliation based on the similarities with the specific tribal vocabularies.

96

P. A. Gloor et al.

3 Tribefinder in Action In this section, we provide some examples of the use of the Tribefinder system, empirically validating its accuracy. As mentioned above, we are confident that firms may find Tribefinder useful for marketing. While a firm a priori knows what kind of customers it wants to reach through its marketing activities, the ex post results may not be as expected. Tribefinder thus offers a simple instrument to assess the alignment between the expected and actual characteristics of a brand’s virtual tribe, which identifies the network of heterogeneous Twitter users that share an interest in this specific brand. In this way, it might reveal that particular tribes have become (unintentionally) attracted by marketing actions, which may in turn become a possible source of innovation for the firm. To empirically test Tribefinder’s accuracy, we selected four firms, brands, or key individuals (hereafter: brands) for each tribe category, whose target customers’ (or audience’s) characteristics fit with those of the tribes. We then identified and analyzed, using Tribefinder, the tribes of the users that tweeted about these brands, to measure their tribal affiliation and verify its congruence with the brand image. The results are presented below, divided into the three tribal macro-categories.

3.1

Alternative Reality

In this section, we provide the percentage tribal affiliations for brands that specifically target fatherlanders (i.e., CNN, Fox News, MSNBC News, Politico), nerds (i.e., Apple, Microsoft, SpaceX, Star Wars), spiritualists (i.e., Dalai Lama, Paolo Coelho, Osho, YogaWorks), and tree huggers (i.e., Greenpeace, Patagonia, PETA, WWF). On the vertical axis, the percentage of analyzed Twitter users that fall into the specific tribe is reported. Figure 2 shows that the majority of the Twitter users in the virtual tribes of the four selected fatherlander brands correctly fall into the fatherlander tribe. Moreover, as individuals typically belong to several tribes (Bauman 1990), looking at other tribal macro-categories, these users are also sedentary or vegan (depending on the brand) and interested in art (e.g., those tweeting about MSNBC News). The correct functioning of the Tribefinder system becomes even clearer when looking at nerd brands (Fig. 3), which mostly attract nerd individuals. The same holds for spiritualist brands (Fig. 4), whose Twitter users are spiritualist as well. Confirming the validity of our system, Dalai Lama-related individuals properly fall into the vegan tribe, while those associated with YogaWorks also belong to the fitness tribe. For tree hugger brands (Fig. 5), the corresponding Twitter users are accurately classified as tree huggers. Finally, Fig. 6 provides a concise view of the results presented above. Specifically, it shows the tribal affiliations of the aforementioned brands’ virtual tribes only referring to the tribal macro-category of the analysis (i.e., alternative reality). From

Identifying Tribes on Twitter Through Shared Context

97

Fig. 2 Tribal affiliations of Twitter users in the virtual tribes of four fatherlander brands

Fig. 3 Tribal affiliations of Twitter users in the virtual tribes of four nerd brands

Fig. 6, a direct correspondence between brands’ types and individuals’ tribal affiliations is clearly visible. For instance, nerd brands attract nerd Twitter users.

3.2

Lifestyle

In this section we present the average tribal affiliations of the Twitter users engaged with brands that specifically target different lifestyles: fitness (i.e., Adidas, CrossFit,

98

P. A. Gloor et al.

Fig. 4 Tribal affiliations of Twitter users in the virtual tribes of four spiritualist brands

Fig. 5 Tribal affiliations of Twitter users in the virtual tribes of four tree hugger brands

Nike, Peloton), sedentary (i.e., Grubhub, Instacart, Pizza Hut, Seamless), vegan (i.e., Beyond Meat, Impossible Foods, PETA, WWF), and yolo (i.e., Alpinestars, GoPro, Monster Energy, Rockstar Energy). Figure 7 shows that Twitter users in the virtual tribes of fitness brands properly fall into the fitness tribe; this is especially true for those related to the CrossFit brand. These brands also coherently attract individuals belonging to the sport tribe. Moreover, it is worth noting that Nike users are also nerds and fashion individuals. Regarding sedentary brands (Fig. 8), among the tribes in the tribal macro-category of lifestyle, their Twitter users are on average categorized as sedentary. Nevertheless, the strongest classifications emerge when looking at the tribal macro-categories of alternative reality and recreation. For instance, the great majority of users tweeting

Identifying Tribes on Twitter Through Shared Context

99

Fig. 6 Alternative reality tribal affiliations of Twitter users belonging to alternative reality brands’ virtual tribes

Fig. 7 Tribal affiliations of Twitter users in the virtual tribes of four fitness brands

about the brands Grubhub, Pizza Hut, and Seamless are nerds; at the same time, those interested in Grubhub and Instacart belong to the travel tribe, while those related to Pizza Hut and Seamless associate with the art tribe. The classification of the individuals tweeting on the four vegan brands (Fig. 9) is in line with the characteristics of these brands, and the same holds true for yolo brands (Fig. 10). Specifically referring to the latter, other relevant tribal affiliations emerge. For instance, GoPro Twitter users are also nerds and interested in travels, while individuals tweeting about Alpinestar, Monster Energy, and Rockstar Energy clearly fall into the sport tribe. In Fig. 11 the tribal affiliations of the selected brands’ tribal macro-categories are shown (i.e., lifestyle). Figure 11 clearly shows that brands succeed in attracting

100

P. A. Gloor et al.

Fig. 8 Tribal affiliations of Twitter users in the virtual tribes of four sedentary brands

Fig. 9 Tribal affiliations of Twitter users in the virtual tribes of four vegan brands

Twitter users belonging to the tribe that best represents the brand (e.g., fitness individuals tweet on fitness brands).

3.3

Recreation

Finally, in this section we provide the same analyses for recreation-oriented tribes. In this case, we selected brands specifically targeting the following recreational activities: art (i.e., Guggenheim, Metropolitan Museum, Museum of Modern Art,

Identifying Tribes on Twitter Through Shared Context

101

Fig. 10 Tribal affiliations of Twitter users in the virtual tribes of four yolo brands

Fig. 11 Lifestyle tribal affiliations of Twitter users belonging to lifestyle brands’ virtual tribes

Smithsonian), fashion (i.e., Chanel, Dior, Gucci, Louis Vuitton), sport (i.e., Broncos, Chicago Bulls, NASCAR, National Football League), and travel (i.e., Delta, Lonely Planet, National Geographic, Southwest). Regarding art-related brands (Fig. 12), while the majority of Twitter users appears to belong to the corresponding tribe, results are not as clear cut as for other brand categories (with the exception of the sports tribe that is significantly less well represented). However, this result is reasonable as the brands we choose are likely to attract travelling individuals, who are also interested in fashion. A more clear classification emerges when analyzing fashion (Fig. 13) and sport (Fig. 14) brands. The fashion tribe affiliation indeed predominates among individuals tweeting about fashion brands; a case in point is Dior; indeed 89% of individuals tweeting about Dior belong to the fashion tribe. The same trend exists for sport

102

P. A. Gloor et al.

Fig. 12 Tribal affiliations of Twitter users in the virtual tribes of four art brands

Fig. 13 Tribal affiliations of Twitter users in the virtual tribes of four fashion brands

brands, as the dominant tribal affiliation in the macro-category of recreation is sport. In this case, also the affiliations regarding the other two tribal macro-categories seem to be reasonable; for instance, users that tweet about Chicago Bulls are mainly sedentary and nerd individuals. Also for travel-related brands, the Tribefinder system works well (Fig. 15) as the majority of Twitter users tweeting about these brands are classified as members of the travel tribe. Similar to the previous tribal macro-categories, Fig. 16 shows a synthesis of the results with reference to recreation tribes. We find again good correspondence between the type of brand and the tribal affiliations of the individuals tweeting about the brand. The clearest results are those regarding fashion and sport brands, where the great majority of Twitter users fall into the fashion and sport tribe,

Identifying Tribes on Twitter Through Shared Context

103

Fig. 14 Tribal affiliations of Twitter users in the virtual tribes of four sport brands

Fig. 15 Tribal affiliations of Twitter users in the virtual tribes of four travel brands

respectively. The classification is somewhat less clear for art- and travel-related brands; this result likely depends on the brands’ characteristics.

4 Further Validation of Tribefinder Results Section 3 presented an intuitive validation of the Tribefinder results. To additionally verify the accuracy of our classification algorithm, two independent annotators manually assigned tribal affiliations to 500 Twitter users randomly extracted from

104

P. A. Gloor et al.

Fig. 16 Recreation tribal affiliations of Twitter users belonging to recreation brands’ virtual tribes Table 2 Tribefinder classification accuracy Tribal macro-category Alternative reality Lifestyle Recreation

Classification accuracy 81.2% 68.8% 69.8%

Kappa statistic 0.731 0.573 0.580

a generic database of tweets covering different topics. The inter-rater agreement between their independent classifications, measured by means of Cohen’s kappa, was high (greater than 0.80). The two annotators then met to find an agreement on discordant cases. Their tribe allocations were subsequently matched with those produced by Tribefinder. The analysis of confusion matrices produced good results in terms of accuracy and Kappa statistic (see Table 2).

5 t-SNE Visualization of Tribe Members As a last step in our analysis, we present the t-SNE visualization approach, which can also be used to have an idea of the quality of the tribe classification generated by the Tribefinder system. t-SNE (Maaten and Hinton 2008) is a popular method for visualizing high-dimensional data. In contrast to classic dimensionality reduction methods like PCA (Jolliffe 2011), which are mainly concerned with preserving large pairwise distances between the data points, t-SNE successfully captures much of the local structure of the high-dimensional representation while also giving an idea about the global structure such as the existence of clusters (Maaten and Hinton 2008). Using this technique, it is possible to visualize the members of the tribes and visually inspect the quality of their tribe assignments by verifying if the individual tribe members are clustered closely together.

Identifying Tribes on Twitter Through Shared Context

105

We illustrate the power of our approach by adding new tribes to the ones illustrated in previous sections (shown in Sects. 5.4–5.6) and visually inspecting if they form cohesive clusters. New tribes can be created by selecting hundreds of “tribe leaders” for a topic and computing their common vocabulary through deep learning (De Oliveira and Gloor 2018). In our approach we selected a subset of tribes—according to the macro-categories or by individually selecting some of the tribes—for which we wanted a t-SNE visualization. We gathered all the twitter users associated with the tribes. The next step was to fetch the 200 most recent tweets of each of the gathered twitter users and tokenize the tweets content for further processing. The tokenization included getting rid of stop words, interpunctuation, URLs, unnecessary whitespace, and tokens which were too short. Having the tokens for each user and thereby also the tokens of the tribes, which is just the collection of all the tokens of its members, we calculated the tf-idf scores (Salton and McGill 1986) for the tokens for the individual users as well as for the tribes. At the same time, we calculated the unigram probability, i.e., the word probability, of all the tokens that we have encountered. The tf-idf scores can be used to define and restrict the vocabulary for further analysis: we selected from each tribe 200 distinct tokens sorted by their importance according to the tribe’s tf-idf. The constructed vocabulary defined which tokens have been then embedded to generate user vectors. We used a pretrained fastText (Bojanowski et al. 2016) word embedding to embed the individual tokens into 300 dimensional vectors. An advantage of using fastText embeddings is its capability to use subword information. This allowed us to obtain embedding vectors for compound words, which in other cases are often not part of the embedding vocabulary. An example of such compound words is hashtags (e.g., #photooftheday), which often contain useful information about a tweet. Using the embedding we got a collection of vectors for each user. Our goal was to represent each user with a single vector, which can be used for the visualization with t-SNE. To combine the collection of vectors into a single representative vector, we tried different methods like, for instance, summing and weighted average (White et al. 2015) using td-idf scores. The best results were achieved with the approach described in the work by Arora et al. (2016). The idea was to first aggregate the vectors by weighting them with their corresponding unigram probability and then summing them up. We then ended up with a single vector for each user. We stacked these vectors together into a matrix and factorized it using SVD. This gave us the eigenvectors of our user vector matrix. We then proceeded by subtracting from each user vector the first eigenvector. Intuitively we can think of subtracting the most common properties of all the users of the tribes. After this step we used t-SNE to reduce the dimensionality of the vectors down to only two dimensions in order to visualize them in scatter plots. We then plotted each of the two-dimensional user vectors and colored them according to their assigned tribe. It is important to note that in the steps described above, we only used the tribe information of a user at the end to assign the respective color in the scatter plot.

106

P. A. Gloor et al.

Fig. 17 t-SNE visualization for the alternative reality macro-category

This gave us a visual indication if the tribe assignments of the individual users make sense, i.e., if they seem to cluster and are distinguishable from members of other tribes. This visualization also allowed to identify outliers in the tribes. The information can help to improve the tribes and make them more distinct from each other. In the next subsections, we show some visualization results for the already existing macro-categories as well as for some individually defined subsets of tribes.

5.1

Alternative Reality

The plot shows that the four tribes cluster nicely. The cluster of the tribe fatherlander seems to be a little more separate compared to the other tribes (Fig. 17).

5.2

Lifestyle

In the plot below, we see that the tribes do not seem to be orthogonal to each other. Intuitively this makes sense since, for instance, there are Fitness Youtubers, who promote the vegan lifestyle and would therefore fit into the category Fitness as well as category Vegan (Fig. 18).

Identifying Tribes on Twitter Through Shared Context

107

Fig. 18 t-SNE visualization for the lifestyle macro-category

5.3

Recreation

In Fig. 19, it is possible to easily identify clusters for the tribes sport, travel, and fashion. The tribe art seems instead to be scattered among the others. In Sects. 5.4–5.6, we illustrate the expressive power of the t-SNE algorithm with new tribes not yet included in the previous analysis.

5.4

Ideology

In this plot, the tribes socialism and liberalism are nicely clustered. The tribes capitalism and complainers both have sections where their members cluster, but in the middle of the plot, there is some overlap with members of other tribes (Fig. 20).

5.5

LGBT vs. Anti-LGBT

In this case, where we have two tribes opposing each other by their topic, we get for the most part a clear distinction. There are some members of both tribes, which do not seem to communicate the same way as their more aligned peers (Fig. 21).

108

Fig. 19 t-SNE visualization for the recreation macro-category

Fig. 20 t-SNE visualization for the ideology macro-category

P. A. Gloor et al.

Identifying Tribes on Twitter Through Shared Context

109

Fig. 21 t-SNE visualization for the custom tribes LGBT vs. anti-LGBT

5.6

Journalist vs. Politician

In general, we get a nice separation between the two large tribes. On both sides there are some individual members which are in the cluster of the respective opposite tribe. For the politician tribe, there seem to be two clusters, the big one in the lower middle as well as the smaller one in the upper half on the left side (Fig. 22).

6 Conclusion In this paper we introduce Tribefinder, a novel system that is able to identify tribal affiliations of Twitter users. Leveraging tribal vocabularies, it analyzes an individual’s words used on Twitter and categorizes her/him into tribes. We present its functionality for three specific tribal macro-categories (alternative reality, lifestyle, and recreation), which are taken as examples. Tribefinder can be easily extended to alternative tribal macro-categories depending on users’ needs. We are convinced that this system will be of value for both researchers and firms. The advent of the Internet and the diffusion of social networking platforms changed marketing paradigms (Burton and Soboleva 2011), and scholars are more and more advising firms to get rid of traditional marketing strategies (Addis and Podesta 2005; Canniford 2011) and to look for new solutions able to incorporate the essence of the

110

P. A. Gloor et al.

Fig. 22 t-SNE visualization for the custom tribes journalist vs. politician

tribes interested in the products or services they offer (Cova and Cova 2002; Moutinho et al. 2007). Tribe characteristics may indeed affect the success of both a marketing campaign and the firm itself (e.g., Holzweber et al. 2015). Overcoming the limits of traditional methodologies that have been used in the past to study tribes, Tribefinder allows scholars and practitioners to easily identify Twitter users’ tribal affiliations and have a clear picture of their characteristics. The information gathered through this system thus potentially constitutes a foundation for future research—e.g., understanding how firms may rely on tribes as a strategic resource (Cova and Cova 2002)—as well as for firms to develop a better understanding of their brand’s virtual tribes on Twitter, to measure the efficiency of their marketing campaigns, and to set up or adjust their marketing strategies. Acknowledgments We are grateful to Fabrizio Marini, for his help in data collection and tribe annotation.

References Addis, M., & Podesta, S. (2005). Long life to marketing research: A postmodern view. European Journal of Marketing, 39, 386–413. Arora S, Liang Y, Ma T (2016) A simple but tough-to-beat baseline for sentence embeddings Bauman, Z. (1990). Thinking sociologically. Oxford: Blackwell.

Identifying Tribes on Twitter Through Shared Context

111

Bojanowski P, Grave E, Joulin A, Mikolov T (2016) Enriching word vectors with subword information. Bringay, S., Béchet, N., Bouillot, F., Poncelet, P., Roche, M., & Teisseire, M. (2011). Towards an on-line analysis of tweets processing. Database and expert systems applications. Heidelberg: Springer. Burton, S., & Soboleva, A. (2011). Interactive or reactive? Marketing with Twitter. Journal of Consumer Marketing, 28, 491–499. Canniford, R. (2011). How to manage consumer tribes. Journal of Strategic Marketing, 19, 591–606. Cova, B. (1996). What postmodernism means to marketing managers. European Management Journal, 14, 494–499. Cova, B., & Cova, V. (2002). Tribal marketing: The tribalisation of society and its impact on the conduct of marketing. European Journal of Marketing, 36, 595–620. Cova, B., & Pace, S. (2006). Brand community of convenience products: New forms of customer empowerment–the case “my Nutella The Community”. European Journal of Marketing, 40, 1087–1105. Cova, B., & White, T. (2010). Counter-brand and alter-brand communities: The impact of Web 2.0 on tribal marketing approaches. Journal of Marketing Management, 26, 256–270. De Oliveira, J. M., & Gloor, P. A. (2018). GalaxyScope: Finding the “Truth of Tribes” on social media. Collaborative innovation networks. Cham: Springer. Dionísio, P., Leal, C., & Moutinho, L. (2008). Fandom affiliation and tribal behaviour: A sports marketing application. Qualitative Market Research: An International Journal, 11, 17–39. Garry, T., Broderick, A. J., & Lahiffe, K. (2008). Tribal motivation in sponsorship and its influence on sponsor relationship development and corporate identity. Journal of Marketing Management, 24, 959–977. Goulding, C., Shankar, A., & Canniford, R. (2013). Learning to be tribal: Facilitating the formation of consumer tribes. European Journal of Marketing, 47, 813–832. Hamilton, K., & Hewer, P. (2010). Tribal mattering spaces: Social-networking sites, celebrity affiliations, and tribal innovations. Journal of Marketing Management, 26, 271–289. Holzweber, M., Mattsson, J., & Standing, C. (2015). Entrepreneurial business development through building tribes. Journal of Strategic Marketing, 23, 563–578. Jolliffe, I. (2011). Principal component analysis. In International encyclopedia of statistical science (pp. 1094–1096). Berlin: Springer. Kozinets, R. V. (1999). E-tribalized marketing? The strategic implications of virtual communities of consumption. European Journal of Marketing, 17, 252–264. Maaten, L. V. D., & Hinton, G. (2008). Visualizing data using t-SNE. Journal of Machine Learning Research, 9, 2579–2605. Maffesoli, M. (1996). The time of the tribes. London: Sage. Moutinho, L., Dionísio, P., & Leal, C. (2007). Surf tribal behaviour: A sports marketing application. Marketing Intelligence & Planning, 25, 668–690. Richardson, B. (2013). Tribal marketing, tribal branding: An expert guide to the brand co-creation process. New York: Springer. Salton G, McGill MJ (1986) Introduction to modern information retrieval. Taute, H. A., & Sierra, J. (2014). Brand tribalism: An anthropological perspective. Journal of Product & Brand Management, 23, 2–15. White, L., Togneri, R., Liu, W., & Bennamoun, M. (2015). How well sentence embeddings capture meaning. 20th Australasian Document Computing Symposium (p. 9).

Part III

Healthcare Applications

Social Media Teams of Hospitals as Mediators in Digital Health Ecosystems Michael Beier and Sebastian Früh

Abstract In this paper, we elaborate relevant theoretical and empirical foundations from the literature and prepare a conceptual framework on how social media teams of hospitals can act as mediators of various parties in their digital health ecosystem. Furthermore, we present our analytical approach as well as preliminary findings from our ongoing research. The paper is a research in progress from the second phase of a project on strategic use of social media in hospitals, which started in April 2018. Its regional focus is on Austria, Germany, and Switzerland, where we conduct empirical studies on the research questions.

1 Introduction Social media platforms provide important communication and marketing channels for hospitals (Smith 2017). They are a valuable tool to communicate with specific segments of the general population (Campbell et al. 2014). In some contexts, people perceive social media presence of a hospital as signal for higher service quality (McCaughey et al. 2014). Furthermore, organizations can use social media to learn more about their various stakeholder groups (Beier 2016; Stieglitz et al. 2014). Therefore, hospitals could apply social media platforms as strategic communication and relationship tools in a complex environment of various stakeholders. However, current state of research is that most hospitals actually use their official social media accounts mainly as a general marketing and public relations tool (e.g., Kordzadeh and Young 2015; Wong et al. 2016). Like many other organizations, hospitals still show significant challenges to apply social media in a strategic manner to pursue concrete purposes for their organization (except for general brand marketing or public relations) or the surrounding digital health ecosystem (Beier and Wagner

M. Beier (*) · S. Früh University of Applied Sciences HTW Chur, Swiss Institute for Entrepreneurship, Chur, Switzerland e-mail: [email protected]; [email protected] © Springer Nature Switzerland AG 2019 Y. Song et al. (eds.), Collaborative Innovation Networks, Studies on Entrepreneurship, Structural Change and Industrial Dynamics, https://doi.org/10.1007/978-3-030-17238-1_6

115

116

M. Beier and S. Früh

2016). However, we assume significant potentials for enhanced use of social media platforms by hospitals. As a key for such advancements, we consider it necessary, to enhance the perspective on hospitals’ social media teams. In this paper, we elaborate theoretical and empirical foundations from the literature and prepare a conceptual framework on how social media teams of hospitals can act as mediators of various parties in their surrounding digital health ecosystem. Furthermore, we present how we apply this new perspective in our research approach, and we present preliminary findings from our ongoing research. The paper is on research in progress from the second phase of a project on strategic use of social media in hospitals, which started in April 2018. Its regional focus is on Austria, Germany, and Switzerland, where we currently conduct several empirical studies on the research questions. In the overall project, we mainly focus on social media platforms where hospitals can run own accounts. This refers primarily to general-purpose social networks (e.g., Facebook), business-related social networks (e.g., LinkedIn), and content-sharing platforms (e.g., YouTube). In contrast, we do not consider social media platforms where users can communicate about hospitals but which do not give hospitals the opportunity to maintain an own account, like platforms where employees (e.g., Glassdoor) or patients (e.g., whichhospital.ch) can review hospitals (DeKay 2013). Furthermore, we also do not include social media platforms, which are mainly used on a project basis but not for continuous social media presence such as crowdfunding platforms (Beier and Wagner 2015). In the first phase of the project, we have analyzed to what extent Swiss hospitals use different social media platforms. Our results show that Facebook (67% of hospitals are present), YouTube (55%), and LinkedIn (53%) are the platforms most commonly used (Beier and Früh 2018). However, hospitals running own accounts on social media platforms differ considerably in their actual usage and communication activities. In the second phase of our project, we try to understand in detail the structures and processes behind hospitals’ adoption and usage of social media platforms as well as how social media activities of hospitals can be improved in relation to specific purposes. Therefore, in this phase of our research project, we focus on the following questions: • How are formal structures and processes of social media platform usage designed in hospitals? • What informal structures and processes complement these? • What are the potentials and challenges of social media teams in hospitals to mediate between various parties inside and outside the hospital?

2 Social Media and Hospitals Social media represent all types of mobile- and web-based applications that allow individual or collective users to create, share, and modify user-generated content through highly interactive platforms (Kaplan and Haenlein 2010). Such informal

Social Media Teams of Hospitals as Mediators in Digital Health Ecosystems

117

platforms provide features to all participants, to generate profiles, connect with others, upload content and other files, as well as to see, read, and comment content and profiles of others (Beier and Wagner 2016). Social media started with high expectations on potential enhancements in health communication. However, until now several limitations of social media applications (e.g., quality concerns, information overload, assumed inappropriateness) limit an extensive realization of concrete benefits in the health sector (Moorhead et al. 2013). Against this background, social media teams of hospitals seem a valuable communicator in digital ecosystems. Hospitals are well-established institutions within a health ecosystem and interact with many parties in the field on a regular basis. However, until now many hospitals use their official social media channels mainly for general marketing or public relations (e.g., Kordzadeh and Young 2015; Wong et al. 2016). Like many other organizations (except applications in brand marketing), hospitals still have considerable difficulties to apply social media in a strategic manner to pursue concrete purposes for their organization or the surrounding health ecosystem (Beier and Wagner 2016). In a similar manner, research on social media applications by hospitals, so far, remains at a rather superficial level (Moorhead et al. 2013). Mainly applying open accessible data of hospitals’ social media channels, many studies focus on simple adoption, usage, and engagement metrics as well as hospital characteristics influencing these (e.g., Beier and Früh 2018; Griffis et al. 2014; Martinez-Millana et al. 2017; Richter et al. 2014). Until now, only few studies apply qualitative and mixed methods to analyze how hospitals use their social media channels in detail. For instance, one study observes that US children’s hospitals post 35% general health information and 35% commercials for their services in their social media channels (Wong et al. 2016). A qualitative analysis of hospitals’ social media posts finds that hospitals succeed rather badly to follow their own content strategies and that they mainly post superficial trivia and commercials for their own services (Kordzadeh and Young 2015). Social media recruitment and employer branding of hospitals is one of few concrete application purposes researchers already investigated more extensively (e.g., Carpentier et al. 2017). Overall, more studies analyze risks of hospitals’ social media activities than studies investigate potentials and difficulties of their strategic use (Richter et al. 2014). On the one hand, this could be due to the generally high requirements and regulations on security, privacy, and data interchange in the health sector. On the other hand, hospitals may lack clear strategies to generate concrete benefits out of their official social media applications. Without a clear understanding and a certain expectancy of such benefits, hospitals are not motivated to apply their official social media channels seriously as means of substantial stakeholder interaction or service provision (Bermúdez-Tamayo et al. 2013). In contrast, they often may participate rather symbolically as they assume own presences in new social media platforms as signal of their innovativeness for the broad public (Beier et al. 2013; McCaughey et al. 2014). One reason for the (practical as well as academic) difficulties described above could be a widespread simplistic perspective on social media as a general branding

118

M. Beier and S. Früh

or public relations tool (Bruhn et al. 2012; Henry and Harte 2012). Therefore, regarding social media communication, a focal organization often is perceived as one homogeneous entity and its external audience as a homogeneous or unspecified population. In a similar manner, many studies on social media applications of organizations apply the TOE framework (e.g., Mousavi and Demirkan 2013; Panagiotopoulos and Barnett 2015; Sharif et al. 2015). This framework, originally developed by Tornatzky and Fleischer (1990) to understand technological innovations, provides three separate context dimensions (TOE: technological, organizational, and environmental context) to systematically analyze organizations and their adaptation and usage of a new information technology. However, concerning social media (especially with regard to hospitals), organizational boundaries are fuzzy and complex networks of internal and external stakeholders have to be taken into account (McFarland and Ployhart 2015). Therefore, it seems analytically not well suited in an empirical research approach to separate an organization, its environment, and social media (as the respective technology) strictly from each other. Thus, in our project, we seek to apply a more comprehensive perspective, which allows us to cover and integrate fuzzy organizational boundaries as well as complex network structures into our further investigation. Especially we are interested in the interplay of motivations, behaviors, and communications between individual and collective levels on a formal (e.g., teams, departments, or organizations) and an informal (e.g., shared meanings, social identities, cognitive distances) basis. For this purpose, in the next section, we evaluate a general perspective of digital health ecosystems in combination with basic ideas of COINs as particularly applicable for our project.

3 Digital Health Ecosystems A digital ecosystem is “. . . an open, loosely coupled, domain clustered, demanddriven, self-organizing agents’ environment, where each specie is proactive and responsive for its own benefit or profit” (Chang and West 2006, p. 6). With their characteristics of openness, self-organization, and distributed swarms of autonomous agents, digital ecosystems obtain several similarities to Collaborative Innovation Networks (COINs). However, there also are significant differences: COINs are defined as “virtual communities interacting on a global scale . . . made up of self motivated people who share a common vision, meeting on the web to exchange ideas, knowledge, experiences and to work in a collaborative way to achieve a common goal” (De Maggio et al. 2009, p. 6). Although, digital ecosystems are open and selforganized communities, which are mediated and complemented by digital applications, they do not necessarily have to be coordinated by a shared vision or a common goal (Dong et al. 2011). In fact, digital ecosystems are much more divided and diverse in the purposes different parties pursue. Whereas in COINs the focal purpose is to collaborate on a new idea or an innovation (Gloor et al. 2003), all agents in digital ecosystems are free to pursue their individual objectives including strategic positioning or sheer profit maximization (Chang and West 2006).

Social Media Teams of Hospitals as Mediators in Digital Health Ecosystems

119

In recent years, concepts of digital ecosystems have been applied increasingly to the health sector (Dong et al. 2011). Digital health ecosystems have some specific characteristics enhancing the general concepts. First, in many fields of the health sector, digital health ecosystems have a strong focus on IT systems (hard- and software) and their interoperability. Topics like electronic health records and integrated healthcare information systems shifted the focus mainly on health and medical informatics as well as IT architectures (Serbanati et al. 2011). This is especially true for hospitals with their high investments in information technology and their multitudes of running applications (Mettler and Pinto 2018). In contrast, more fundamental perspectives of social networks, societal ecosystems, and different domain cultures have received far less attention (Beier and Semrau 2008). Second, the health sector is specific in its very high requirements on security and privacy. Compliance officers as well as most members of organizations in the health sector pay particular attention to privacy and data security laws and regulations (Warkentin et al. 2011). Therefore, a special focus in digital health ecosystems lies on the security of data and information exchange in its networks (Iyawa et al. 2016). Correspondingly, in digital health ecosystems, the openness of communication and interaction as known from collaborative networks as well as from digital ecosystems in general is considerably limited in many ways. Third, several regional aspects influence and structure digital health ecosystems. For instance, national or multinational institutions design the general health system and its regulations (Thomson et al. 2015). Furthermore, participation, interaction, and information behavior of agents in digital health ecosystems are influenced by national and regional culture, for instance, in the context of health literacy (Batterham et al. 2016). In addition, many medical services and institutions have a regional focus in their service provision. In this regard, hospitals obtain a central role in their respective digital health ecosystems (Serbanati et al. 2011). Against this background, we understand digital health ecosystems as loosely coupled conglomerates of various domain clusters, in which some of them act as COINs working collectively on a new idea or pursuing to generate radically new innovation (Gloor et al. 2003), whereas other domain clusters and agents follow different objectives or individual purposes in their activities. Within these digital health ecosystems, hospitals are connected to many stakeholders of various parties on a regular but highly formalized basis.

4 Analytical Approach and Preliminary Findings Inspired by this frame of reference, we designed a qualitative study approach to analyze how the complex structures and processes of hospitals and their environment manifest in their social media activities. As outlined above, in our study it is particularly important that we capture adequately relevant internal structures, interactions, domains, and social identities within hospitals. Therefore, we decided to interview individuals on different levels and departments within hospitals. Because

120

M. Beier and S. Früh

Table 1 Comparison of cases Case 1 One hospital with several locations Early adopter Heavy use of social media Public relations logic

Case 2 Group of hospitals Still not adopter Evaluating potential applications Business development logic

of the complexity of structures and processes investigated within the hospitals, we are only able to include very few cases into our study. Thus, we decided to analyze, two very different cases from Switzerland that are opposite in terms of several dimensions. In a first round, we conducted four (Case 1) and seven (Case 2) interviews with members of the organizations from different levels and departments. Table 1 shows the main differences between both cases: Case 1 is an independent hospital with several locations. It is already heavily using various social media platforms and publishing frequently new content on a regular basis. However, the hospital is using social media as a general public relations tool what makes it overall easier for them. They run social media mainly like an online news publisher, generating all posts by a content production team coordinated by an editorial plan. However, like social media teams in other fields, they also have established several internal processes to be able to react on posts from users in their channels in a timely manner. The social media team relies on a close relationship network within the hospital to gather well-informed inputs for their social media conversations with users in the channels. Because they run their social media accounts in a healthcare institution, they have to follow carefully legal requirements on privacy and medical conversations (that originally were not designed for social media platforms). Case 2 is a small group of hospitals, which still did not start social media. They want to follow a business development logic with their own presence in social media platforms. Therefore, they only want to start social media activities if they can clearly define in advance what concrete purposes they pursue with these (meaning purposes that go beyond symbolic social media presences). Furthermore, they see the necessity to engage their own employees into the social media activities of the organization. On the one hand, for them it is important that the social media activities of the hospitals have broad support from their employees and other internal stakeholders. On the other hand, they seek to reduce the risks of inappropriate communications within their social media channels. Therefore, they are now developing in advance a guiding social media framework to influence how employees as well as other stakeholders use their future social media channels. Although the channels will be run following a participatory approach, it is planned that a core social media team provides own content as well as moderates and governs the activities of all other stakeholders. Both cases show boundary spanning and mediating activities of social media teams in the hospitals. As the social media platforms themselves provide the interface to external stakeholders of the organizations, the hospitals’ main focus is

Social Media Teams of Hospitals as Mediators in Digital Health Ecosystems

121

Fig. 1 Boundary spanning network structure of social media teams

actually on moderation and gatekeeping of internal stakeholders’ participation. However, our interviews also indicated several external stakeholder groups, which have different importance for different groups within the hospitals. First, hospitals are geared to attract (specific) patients for their service offerings. Second, they depend on access to well-educated further employees. Third, they seek to establish and maintain good relationships to relevant health professionals in their field. Furthermore, hospitals are overall dependent on legitimacy in the general population and in politics. Because of their position in the processes of social media communication, social media teams of a focal hospital are in a boundary spanning position between the internal and external stakeholder groups (see Fig. 1). As described above, current research provides first fundamentals on social media communication of hospitals with various groups of external stakeholders (e.g., (potential) patients, healthcare professionals, and potential employees). However, most empirical findings show that hospitals mainly communicate only unidirectional with these groups (Vanzetta et al. 2014). Cases on how hospitals could use feedback from social media channels are already known from network analyses in medical research (e.g., Gloor et al. 2016). Such analyses can also provide useful insights relevant for other internal stakeholders in hospitals. However, such measures are mainly implemented only by specific units within hospitals (often for scientific research), but are not organized and integrated on the level of the whole organization. A plausible explanation for this may be that all the internal and external stakeholder groups are too different in their focal perspectives on a hospital. For instance, a hospital manager illustrated us in an interview, how differently various stakeholder groups (e.g., patients, clinicians, nursing stuff, referring doctors, and insurers) could interpret a social media post on cost savings in a specific service offering of the hospital. This variety between stakeholder groups is a major challenge for social media teams in hospitals to mediate conversations successfully. Furthermore, we see a key for successful social media teams in hospitals in the realization of adequate internal processes and structures supporting and complementing

122

M. Beier and S. Früh

external interactions. Research on organizational network development shows that agents’ successful boundary spanning activities with external partners often necessitate sufficient internal relations and support processes within an organization (Beier 2011; Semrau and Beier 2015). In a concrete manner, we already observed this in one hospital in our interviews where the social media team is well connected to a diverse set of internal stakeholders in the hospital (especially administration, doctors, management, and nursing). Only this internal relationship network allows the team to react in a proper and timely manner on external requests via their social media channels. Another field of application in this regard is processes integrating social media analytics into the organization (Beier 2016; Stieglitz et al. 2014). On the one hand, this addresses adequate analytics within the social media teams to optimize the social media communication itself. On the other hand, this refers to strategic implications of social media, where learnings from social media communications are considered in the strategic management of hospitals (Kurniawati et al. 2013). However, research on internal structures and processes to support social media communication is still in its infancy.

5 Conclusion This research-in-progress paper gives a compact overview of our conceptual framework how social media teams of hospitals can act as mediators of various parties in their surrounding digital health ecosystem. In our ongoing empirical research, we particularly focus on the interplay between internal relations (connecting social media teams of hospitals with relevant intraorganizational stakeholders) and their bidirectional interactions with external stakeholders. We hope that a better understanding of internal and external networks of relationships and interactions will help to enhance the benefits of hospitals’ social media activities. In addition, we see such social media practices as a promising opportunity to complement highly formalized information interchange in digital health ecosystems by more social, open, and informal communication between various stakeholders.

References Batterham, R. W., Hawkins, M., Collins, P. A., Buchbinder, R., & Osborne, R. H. (2016). Health literacy: Applying current concepts to improve health services and reduce health inequalities. Public Health, 132, 3–12. Beier, M. (2011). Die Entwicklung Sozialer Netzwerke von Gründerteams. Dissertation Thesis, Universität zu Köln Beier, M. (2016). Startups’ experimental development of digital marketing activities. A case of online-videos. Social Science Research Network (SSRN) Electronic Journal. http://ssrn.com/ abstract¼2868449 Beier, M., & Früh, S. (2018). Nutzung von Social Media Plattformen durch Schweizer Spitäler Eigene Accounts, Nutzungsintensitäten und Reichweiten. Forschungsbericht der Hochschule für Technik und Wirtschaft HTW Chur.

Social Media Teams of Hospitals as Mediators in Digital Health Ecosystems

123

Beier, M., & Semrau, T. (2008). Collaborative networks. A beneficial perspective on health care in developing countries. In P. A. Cholewka & M. M. Motlagh (Eds.), Health capital and sustainable socioeconomic development (pp. 33–48). Boca Raton, FL: Taylor Francis Group. Beier, M., & Wagner, K. (2015). Crowdfunding success: A perspective from social media and ecommerce. In Proceedings of the 36th International Conference on Information Systems (ICIS), Fort Worth, TX. Beier, M., & Wagner, K. (2016). Social media adoption: Barriers to the strategic use of social media in SMEs. In Proceedings of the 24th European Conference on Information Systems (ECIS), Istanbul, Turkey. Beier, M., Früh, S., & Wagner, K. (2013). Social Media Aktivitäten von KMU in der Ostschweiz (Social Media Activities of SMEs in Eastern Switzerland). Social Science Research Network (SSRN) Electronic Journal. http://ssrn.com/abstract¼2423818 Bermúdez-Tamayo, C., Alba-Ruiz, R., Jiménez-Pernett, J., García Gutiérrez, J. F., Traver-Salcedo, V., & Yubraham-Sánchez, D. (2013). Use of social media by Spanish hospitals: Perceptions, difficulties, and success factors. Telemedicine and e-Health, 19(2), 137–145. Bruhn, M., Schoenmueller, V., & Schäfer, D. B. (2012). Are social media replacing traditional media in terms of brand equity creation? Management Research Review, 35(9), 770–790. Campbell, D. A., Lambright, K. T., & Wells, C. J. (2014). Looking for friends, fans, and followers? Social media use in public and nonprofit human services. Public Administration Review, 74(5), 655–663. Carpentier, M., Van Hoye, G., Stockman, S., Schollaert, E., Van Theemsche, B., & Jacobs, G. (2017). Recruiting nurses through social media: Effects on employer brand and attractiveness. Journal of Advanced Nursing, 73(11), 2696–2708. Chang E., & West M. (2006). Digital ecosystems a next generation of the collaborative environment. In Proceedings of the Eighth International Conference on Information Integration and Web-based Applications (WAS), Yogyakarta, pp. 3–23. De Maggio, M., Gloor, P. A., & Passiante, G. (2009). Collaborative innovation networks, virtual communities and geographical clustering. International Journal of Innovation and Regional Development, 1(4), 387–404. DeKay, S. H. (2013). Peering through Glassdoor.com – what social media can tell us about employee satisfaction and engagement. In Conference on Corporate Communication, New York. Dong, H., Hussain, F. K., & Chang, E. (2011). A framework for discovering and classifying ubiquitous services in digital health ecosystems. Journal of Computer and System Sciences, 77(4), 687–704. Gloor, P. A., Laubacher, R., Dynes, S. B., & Zhao, Y. (2003). Visualization of communication patterns in collaborative innovation networks-analysis of some W3C working groups. In Proceedings of the Twelfth International Conference on Information and Knowledge Management (CIKM), New Orleans, pp 56–60. Gloor, P. A., Maddali, H. T., & Margolis, P. A. (2016). Comparing online community structure of patients of chronic diseases. International Journal of Organisational Design and Engineering, 4 (1–2), 113–136. Griffis, H. M., Kilaru, A. S., Werner, R. M., Asch, D. A., Hershey, J. C., Hill, S., et al. (2014). Use of social media across US hospitals: Descriptive analysis of adoption and utilization. Journal of Medical Internet Research, 16(11). Henry, M., & Harte, B. (2012). A new model for assessing the value of social media activity. Journal of Brand Strategy, 1(3), 234–239. Iyawa, G. E., Herselman, M., & Botha, A. (2016). Digital health innovation ecosystems: From systematic literature review to conceptual framework. Procedia Computer Science, 100, 244–252. Kaplan, A. M., & Haenlein, M. (2010). Users of the world, unite! The challenges and opportunities of social media. Business Horizons, 53(1), 59–63.

124

M. Beier and S. Früh

Kordzadeh, N., & Young, D. (2015). Understanding how hospitals use social media: An exploratory study of Facebook posts. In Proceedings of the 21st Americas Conference on Information Systems (AMCIS), Puerto Rico. Kurniawati, K., Shanks, G. G., & Bekmamedova, N. (2013). The business impact of social media analytics. In Proceedings of the 21st European Conference on Information Systems (ECIS), Utrecht, Netherlands. Martinez-Millana, A., Fernandez-Llatas, C., Bilbao, I. B., Salcedo, M. T., & Salcedo, V. T. (2017). Evaluating the social media performance of hospitals in Spain: A longitudinal and comparative study. Journal of Medical Internet Research, 19(5), e181. McCaughey, D., Baumgardner, C., Gaudes, A., LaRochelle, D., Wu, K. J., & Raichura, T. (2014). Best practices in social media: Utilizing a value matrix to assess social media’s impact on health care. Social Science Computer Review, 32(5), 575–589. McFarland, L. A., & Ployhart, R. E. (2015). Social media: A contextual framework to guide research and practice. Journal of Applied Psychology, 100(6), 1653–1677. Mettler, T., & Pinto, R. (2018). Evolutionary paths and influencing factors towards digital maturity: An analysis of the status quo in Swiss hospitals. Technological Forecasting and Social Change, 133, 104–117. Moorhead, S. A., Hazlett, D. E., Harrison, L., Carroll, J. K., Irwin, A., & Hoving, C. (2013). A new dimension of health care: Systematic review of the uses, benefits, and limitations of social media for health communication. Journal of Medical Internet Research, 15(4). Mousavi, S., & Demirkan, H. (2013). The key to social media implementation: Bridging customer relationship management to social media. In Proceedings of the Hawaii International Conference on System Sciences (HICSS), Hawaii, USA. Panagiotopoulos, P., & Barnett, J. (2015). Social media in union communications: An international study with UNI global union affiliates. British Journal of Industrial Relations, 53(3), 508–532. Richter, J. P., Muhlestein, D. B., & Wilks, C. E. (2014). Social media: How hospitals use it, and opportunities for future use. Journal of Healthcare Management, 59(6), 447–461. Semrau, T., & Beier, M. (2015). How specialised and integrated relationship management responsibilities foster new ventures’ network development. International Journal of Entrepreneurial Venturing, 7(1), 47–64. Serbanati, L. D., Ricci, F. L., Mercurio, G., & Vasilateanu, A. (2011). Steps towards a digital health ecosystem. Journal of Biomedical Informatics, 44(4), 621–636. Sharif, M. H. M., Troshani, I., & Davidson, R. (2015). Public sector adoption of social media. Journal of Computer Information Systems, 55(4), 53–61. Smith, K. T. (2017). Hospital marketing and communications via social media. Services Marketing Quarterly, 38(3), 187–201. Stieglitz, S., Dang-Xuan, L., Bruns, A., & Neuberger, C. (2014). Social media analytics. Wirtschaftsinformatik, 56(2), 101–109. Thomson, S., Figueras, J., Evetovits, T., Jowett, M., Mladovsky, P., Maresso, A., et al. (2015). Economic crisis, health systems and health in Europe: Impact and implications for policy. Copenhagen: World Health Organization, Regional Office for Europe. Tornatzky, L. G., & Fleischer, M. (1990). The processes of technological innovation. Issues in organization and management series. Lanham: Lexington Books. Vanzetta, M., Vellone, E., Dal Molin, A., Rocco, G., De Marinis, M. G., & Rosaria, A. (2014). Communication with the public in the health-care system: A descriptive study of the use of social media in local health authorities and public hospitals in Italy. Annali dell’Istituto Superiore di Sanità, 50(2), 163–170. Warkentin, M., Johnston, A. C., & Shropshire, J. (2011). The influence of the informal social learning environment on information privacy policy compliance efficacy and intention. European Journal of Information Systems, 20(3), 267–284. Wong, C. A., Ostapovich, G., Kramer-Golinkoff, E., Griffis, H., Asch, D. A., & Merchant, R. M. (2016). How US children’s hospitals use social media: A mixed methods study. Healthcare, 4 (1), 15–21.

Promoting Holistic Care by Advancing Cultural Competence of Nursing Students in Mainland China Hua Yuan and Caroline Porr

Abstract China is a multicultural society made up of 56 ethnic groups. In addition to the complex cultural healthcare beliefs and traditions, China’s 1.38 billion people live under diverse social and economic conditions. Providing holistic care in such a multifaceted context is challenging for nurses and nursing educators. If nursing students are to provide holistic patient care, then nurse educators must design courses and pedagogic strategies that enhance student capacity to assess and address unique needs including biological, psychological, social, and spiritual problems and concerns. In particular, beginning with spiritual dimension of health is promising because spiritual beliefs are interwoven within the cultural fabric of Chinese people. However, Chinese people do not readily share private thoughts and feelings, and thus even introducing the topic of the provision of spiritual care as part of nursing curricula is challenging. In this article we share the educational content and pedagogical approach that was implemented in an undergraduate multicultural nursing care course that was designed to overcome these challenges. The cultural competency model developed by Camphina-Bacote (Journal of Transcultural Nursing 13:181–184, 2002) was instrumental in structuring learning around spirituality as part of cultural competency training.

1 Introduction Mainland China is a multicultural society. The country has a population of 1.38 billion people distributed among 56 ethnic groupings (Population Reference Bureau 2017). Complex and diverse genetic, linguistic, cultural, and social characteristics

H. Yuan (*) School of Nursing, Jilin University, Changchun, China e-mail: [email protected] C. Porr School of Nursing, Memorial University, St. John’s, Newfoundland, Canada e-mail: [email protected] © Springer Nature Switzerland AG 2019 Y. Song et al. (eds.), Collaborative Innovation Networks, Studies on Entrepreneurship, Structural Change and Industrial Dynamics, https://doi.org/10.1007/978-3-030-17238-1_7

125

126

H. Yuan and C. Porr

create unique health problems which are not seen in the Western world. Long ago, as an ancient civilization, Chinese people had their own religious tradition and philosophy, but during this era of modern economic globalization, China is becoming more of a pluralistic society. In this present day, Chinese people hold rich yet diverse cultural beliefs and values and draw from a variety of spiritual sources. In terms of nursing curricula, to prepare nursing students to provide holistic care to patients, including individuals, families, groups, and communities across the life span and across the continuum of healthcare environments, the question facing nursing educators in Mainland China is “How do we teach nursing students to provide holistic care in such a culturally diverse context?” Holistic care is a philosophy that aspires to ensure patient care which entails consideration of each and every health dimension. The individual patient may have biological, psychological, social, and/or spiritual needs. If nursing students are to provide holistic care, then nurse educators must design courses and pedagogic strategies that enhance student capacity to assess and address biological, psychological, social, and spiritual health needs. However, in the Chinese nursing curriculum, the social and spiritual health dimensions, in particular, are not well understood (Yuan and Porr 2014). Yet, we know from Dr. Narayanasamy’s research that spirituality and cultural beliefs are critical aspects of everyday life (Narayanasamy 2006). Moreover, the inclusion of cultural beliefs in the provision of patient care will enhance quality of care provided and reduce health disparities (AACN 2008). Little is known about educational interventions that may promote holistic nursing care delivery in Mainland China. We believed that broaching spirituality as a central component in a multicultural care course would advance the students holistic nursing care competencies. That is, enabling students to be aware of and value the importance of both culture and spirituality would enable students to, ultimately, address multidimensional health needs consistent with holistic nursing care in Mainland China. The purpose of this article is to present the content and classroom activities that were included in an undergraduate multicultural nursing care course that was implemented in Mainland China and that has effectively heightened spiritual and cultural awareness among undergraduate nursing students.

2 Conceptual Framework Cultural competence is becoming increasingly important in post-secondary education during this era of economic globalization, especially for students enrolled in the health professions. However, according to Blanchet Garneau, the conceptualization of cultural competence is not well understood among health professional educators and students (Garneau and Pepin 2015). Failure to agree on a definition and a model of cultural competence are obstacles to developing, implementing, and evaluating the effectiveness of cultural competence education and training (Alizadeh and Chavan 2016). Camphina-Bacote’s clear and comprehensive cultural competency model (Campinha-Bacote 2002) was chosen to guide the design of the course.

Promoting Holistic Care by Advancing Cultural Competence of Nursing. . .

127

Camphina-Bacote defines cultural competence in his model (The Process of Cultural Competence in the Delivery of Healthcare Services) as “the process in which the healthcare provider continuously strives to achieve the ability to work effectively within the cultural context of a client, individual, family or community” (Campinha-Bacote 2002). The Campinha-Bacote model requires healthcare providers to see themselves as becoming culturally competent rather than already being culturally competent. The integration of cultural competence is a process, not an event, which is comprised of five essential steps that correspond with A.S.K. E.D.: developing Awareness of differences, acquiring Skills to assess those differences, seeking Knowledge of other cultures, Engaging with individuals different from oneself, and having the Desire to change one’s attitudes and views about others (Campinha-Bacote 2002). We adopted the above steps and elements as the main content of the multicultural care course. Specifically, we focused on cultural awareness, cultural knowledge, cultural skills, cultural encounters, and cultural desire.

3 Methods The multicultural care course was offered as an elective course, open to all first-year nursing students in an undergraduate nursing program in Mainland China. The course was developed and taught by an experienced nurse educator who had lived in more than one province in Mainland China and had lived abroad in North America. An outside expert in cultural competency training provided feedback on course concepts and recommended interactive activities designed to engage students and facilitate group learning. The 67 students who enrolled in the class represented 11 of China’s 56 ethnic groups. Students came from 17 provinces and 4 regions of Mainland China. There was 1 male and 66 female students. The average age was 19 years. Class was held twice a week for the month of April. Students received a total of 16 hours of in-class instruction.

3.1

Course Preparation

Given that the conceptualization of cultural competence has been largely based on Western culture and spiritual/religious practices, it was important that we assessed the students’ baseline understanding of culture and spirituality before the commencement of the course. Five questions were posed to students; they were: What is your faith or faith identity? To whom do you turn when you need help? What are your sources of strength and hope? What is spirituality and spiritual care? What is cultural caring? Student responses to the pre-course questions were as follows (Table 1):

128

H. Yuan and C. Porr

Table 1 Student level of understanding of culture (N ¼ 67) Question 1. What is your faith or faith identity?

2. To whom do you turn when you need help? 3. What are your sources of strength and hope?

4. What is spirituality and spiritual care?

5. What is cultural caring?

Response Don’t have faith Buddhist Muslim From Christian family Are curious about Taoism Turn to parents and or friends Do not know or rely on self Believe the sources of strength and hope are from family, especially parental love, support, and expectations Are from themselves, their own dreams and beliefs of a better life in the future Do not know or rely on fate No understanding of the meaning of spirituality and spiritual care Said it is like mental health or psychological health Said it is spirit and mind, feeling, soul, consciousness, hope and strength resources, religion, faith and trust, follow the heart, love; understand, accept and respect the patients; make patients happy, comfortable, peaceful and support; communication Students mentioned that cultural caring is providing care to patients based on their diet, customs, culture, values, faith, nationality, ethnicity, and socioeconomic status

Frequency (%) 87 3 3 3 3 96 4 60

34 6 7 30 63

100

Based on the above pre-course responses, we anticipated that the new course in addition to clarifying the meaning of spirituality would afford students heightened self-awareness of the degree to which their lives were or were not founded on spiritual/religious beliefs. Pedagogically, students first require experience of concepts (e.g., spirituality) before they can fully comprehend a concept (e.g., the spiritual health dimension) and apply knowledge and skills (e.g., spiritual care). Classroom assignments aimed at enhancing self-awareness became key pedagogic strategies.

3.2

The Multicultural Care Course

As before mentioned, cultural awareness, cultural knowledge, cultural skills, cultural encounters, and cultural attitudes were the course foci. The student learning objectives were to enhance insight into personal cultural experiences and spiritual resources. Advance understanding of cultural/spiritual beliefs and values. Acquire cultural/spiritual assessment skills. Gain intercultural communication skills, and develop holistic care plans.

Promoting Holistic Care by Advancing Cultural Competence of Nursing. . .

3.2.1

129

Cultural Awareness

In a relaxed learning environment, students were invited to reflect on the significance of culture for the people of Mainland China. What did they know of the impact of economic globalization and other historical, social, and political influences on Chinese culture? Then in small learning groups, students exchanged personal cultural identities and experiences. Students exchange facilitated cultural self-awareness which is a vital part of cultural competency training that entails self-examination and in-depth exploration of one’s own cultural background, values, and traditions (Harry 1992). The instructor encouraged the students to share their cultural differences and reflect on opposing health beliefs and how they may influence health decisions. The meaning of spirituality was introduced by the instructor, and then in the small learning groups, students exchanged what they knew of Mainland China’s ethnoreligious group affiliations. Self-reflection of individual spiritual journeys and authentic disclosure enabled students to recognize their own spiritual resources and/or identify the spiritual void in their lives. Case studies of patients finding comfort and emotional support when spiritual health is addressed assisted students to value and understand the significance of the spiritual dimension of health and spiritual care for patient healing and recovery.

3.2.2

Cultural Knowledge and Cultural Encounters

Students discussed key concepts such as culture, subcultures, multiculturalism, cultural assimilation, pluralism, cultural acculturation, and cultural competence. Subsequent classroom activities and group assignments and presentations facilitated student application of concepts and relevant knowledge. Assignments included group presentations on Asian, North American, South American, African, Australian, and European cultures. One of the assignments involved group presentations of regional and minority ethno-cultural beliefs found throughout Mainland China followed by individual students sharing emic cultural perspectives. Students were required to listen to the etic cultural perspectives of their peers and then reflect on the emic-etic differences. In doing so, students experienced simulated cultural encounters in the classroom setting allowing nursing students to open themselves up to alternative viewpoints and construct new frames of reference.

3.2.3

Cultural Skills and Cultural Attitudes

Students toward the end of the course focused on the theoretical principles and skills necessary for the delivery of culturally competent care including the establishment of a therapeutic relationship through intercultural communication techniques. The therapeutic relationship is central to all nursing practice (Establishing Therapeutic

130 Table 2 Student enrollment from 2014 to 2017

H. Yuan and C. Porr Course offering date Fall 2014 Spring 2015 Spring 2016 Fall 2017

Number of students enrolled 67 101 121 120

Relationships 2002). Patient-centered, intercultural communication skills equipped students with the ability to know how to tailor their approach, as needed, with culturally diverse patients. The instructor used several case studies of patients with varying cultural characteristics and ethnic beliefs. Students role-played complex scenarios to practice how to communicate and establish therapeutic rapport. In addition, different scenarios depicting both positive and negative role models were presented in front of the class, and students were invited to reflect on their feelings and reactions and identify their cultural attitudes. Although an evaluation of the course was not conducted, student enrollment in the course increased every year, suggesting that students both enjoyed and benefitted from participating in this multicultural care course (Table 2).

4 Discussion and Conclusion Anecdotally, students provided positive valuations after participating in multicultural nursing care course. Moreover, both nursing students and instructors experienced personal growth and professional transformation. Camphina-Bacote’s cultural competency model (ASKED) effectively guided the content and pedagogic strategies of the course. Self-reflection and small learning group discussions enabled students to develop awareness of cultural diversity. Questions posed by the instructor prompted students to think about how cultural diversity is manifest, and at the same time role modeling for students how to conduct a cultural assessment to understand a patient’s belief system. Asking students to research and present the subcultures of Mainland China and the cultures of other countries around the world provided students firsthand encounters with beliefs, values, and traditions different from themselves. Inviting students to reflect on spirituality and their own spiritual health coupled with real-life spiritual care case studies was critical to raising appreciation for the role of the spiritual dimension of health for quality patient care, that is, for holistic nursing care. Skills training was also part of the course which fostered the capacity of students to know how to approach and communicate with patients who may have wear a different cultural lens. Developing the cultural competence of nursing students in Mainland China will undoubtedly ensure that nursing graduates will have the desire to consider the context from which each and every patient derives beliefs about health and healthcare decisions. Given that patients will have been acculturated in any 1 of

Promoting Holistic Care by Advancing Cultural Competence of Nursing. . .

131

the 56 ethnic groups found in Mainland China, cultural awareness and assessment, the knowledge of and respect for diverse beliefs, and seeking to understand spiritual health needs are significant components of holistic nursing care. In addition, in terms of promoting holistic nursing care, we recommend that cultural competence be a key curricular concept that is threaded throughout the baccalaureate nursing program. Students should be building on first-year fundamental knowledge, skills, and attitudes in subsequent classroom, lab, and clinical practice settings. Cultural competence should also be an expected learning outcome during high-fidelity patient simulation. Furthermore, healthcare institutions should be encouraged to provide professional development support and resources to address issues in the workplace that stem from cultural diversity. Practicing nurses should be equipped to reinforce and extend the cultural competency principles that nursing students learn in nursing school.

References AACN – American Association of Colleges of Nursing. (2008). Cultural competency in baccalaureate nursing education. Retrieved July 2011, from http://www.aacn.nche.edu/leading-initia tives/education-esources/competency.pdf (August) Alizadeh, S., & Chavan, M. (2016). Cultural competence dimensions and outcomes: A systematic review of the literature. Health & Social Care in the Community, 24(6), 748–770. Campinha-Bacote, J. (2002). The process of cultural competence in the delivery of healthcare services: A model of care. Journal of Transcultural Nursing, 13(3), 181–184. Establishing Therapeutic Relationships. (2002). http://rnao.ca/bpg/guidelines/establishing-therapeu tic-relationships Garneau, A. B., & Pepin, J. (2015). Cultural competence: A constructivist definition. Journal of Transcultural Nursing, 26(1), 9–15. Harry, B. (1992). Developing cultural self-awareness: The first step in values clarification for early interventionists. Topics in Early Childhood Special Education, 12(3), 333–350. Narayanasamy, A. (2006). The impact of empirical studies of spirituality and culture on nurse education. Journal of Clinical Nursing, 15(7), 840–851. Population Reference Bureau. (2017). 2017 World Population Data Sheet. https://www.prb.org/ 2017-world-population-data-sheet/ Yuan, H., & Porr, C. (2014). Integrating spiritual care into a baccalaureate nursing program in mainland China. Journal of Holistic Nursing, 32(3), 240–243.

Building Shared Environmental Governance for the Future: The Case of a Community COIN Julia C. Gluesing, Ken Riopelle, and Christina Wasson

Abstract This case illustrates the use of mixed methods for analyzing structure, content, and sentiment of a city council task force appointed to guide the city (note, “the city’s” name remains anonymous as per research agreement) in updating their oil and gas drilling ordinance. The novel method approach integrates linguistic conversation analysis and social network influence to analyze the interactions among stakeholders over time as they occur both inside and outside formal meetings. The case study also represents a practical example of conducting e-Research where a small team of four researchers shared a cloud-based PC as a collaborative space for cost efficiency and data sharing and as an educational tool for team members working across mixed PC and Mac operating systems. Although this research is a single case study of one community, the data sources and methods can be scaled to include multiple communities or a larger geographic

This research was sponsored by the US National Science Foundation (NSF), EAGER Grant # BCS-1408169. https://www.nsf.gov/awardsearch/showAward?AWD_ID¼1408169& HistoricalAwards¼false. Last accessed on May 24, 2018. EAGER: Integrating the Analysis of Decision-Making in Meetings with the Analysis of Network Interactions in the Study of Environmental Governance. 2014. Award Number:1408169. https://www.nsf.gov/pubs/policydocs/pappg18_1/pappg_2.jsp#IIE2 Last accessed on May 24, 2018. Note: EAGER means EArly-concept Grants for Exploratory Research. EAGER is a type of NSF proposal used to support exploratory work in its early stages on untested, but potentially transformative, research ideas or approaches. This work may be considered especially “high risk-high payoff” in the sense that it, for example, involves radically different approaches, applies new expertise, or engages novel disciplinary or interdisciplinary perspectives. J. C. Gluesing (*) · K. Riopelle Industrial and Systems Engineering, Wayne State University, Detroit, MI, USA e-mail: [email protected]; [email protected] C. Wasson Department of Anthropology, University of North Texas, Denton, TX, USA e-mail: [email protected] © Springer Nature Switzerland AG 2019 Y. Song et al. (eds.), Collaborative Innovation Networks, Studies on Entrepreneurship, Structural Change and Industrial Dynamics, https://doi.org/10.1007/978-3-030-17238-1_8

133

134

J. C. Gluesing et al.

area that could include a county, state, region, or an entire country. The authors identify traditional methods that can take advantage of data analytics enabled by new technologies.

1 Introduction Environmental governance is increasingly shifting from a top-down approach, in which bureaucrats and scientists make decisions, to a “participatory” approach, in which representatives of diverse stakeholder groups make decisions collaboratively despite the fact they often have competing interests (Buck et al. 2001; Folke et al. 2005; Wondolleck and Yaffee 2000). Implementing the goal of giving all stakeholder groups a voice has proved challenging. Most studies have focused their analysis either on the stakeholder meetings where decisions are negotiated (Dewulf et al. 2011; Dewulf and Bouwen 2012; Roncoli et al. 2011) or on the broader interactions among networks of stakeholders outside of those meetings to influence those meeting decisions (Ostrom 2009; Carlsson and Sandstrom 2008; Lubell 2013; Nagendra and Ostrom 2012). On their own, each of these approaches has serious limitations. The study’s overall objective was to develop a new methodological approach that integrates the analysis of meetings and network interactions. This paper is about the community collaborative innovation network (COIN) that worked to create a new city ordinance to build a shared future for environmental governance. We do not attempt to describe all of the methods, tools, analytical strategies, and findings in the study. The case example illustrates two innovative features that contributed to a comprehensive analysis of multi-stakeholder environmental governance: (1) a novel methodological approach to investigate multi-stakeholder dynamics that enabled us to integrate longitudinal analysis of meeting interactions with the email interactions happening outside the meetings and (2) a practical example of conducting e-Research where a small team of four researchers shared a cloud-based PC as a collaborative space for cost efficiency and data sharing and as an educational tool for team members working across mixed PC and Mac operating systems. We discuss each of these features and conclude with a discussion of future opportunities for researching networked environmental governance.

2 Methodological Integration of Conversation Analysis Data and Email Data Our case example illustrates conversation analysis combined with network analysis of meetings as well as network analysis of email exchanges outside the meetings. Figure 1 is a graphical summary of all of the study’s data sources, study design, methods, and software used in analysis and for the research team to collaborate virtually.

Building Shared Environmental Governance for the Future: The Case of a. . .

135

Fig. 1 Data sources, study design, methods, and software utilized in the study

The study site for our research was a city in the USA whose city council had recently decided to update its ordinance on gas well drilling, because of the rise of hydraulic fracturing in the area. The city council appointed a commission of 5 people to hold a series of 11 public meetings over a 3-month period to make recommendations to the city to revise their current gas well ordinance. We were able to gather the personal email archives of three out of the five commission members. There was enough overlap in the commission practice of copying each other on all commissionrelated email that we were able to create an accurate network of the commission’s email communication over time. (See the footnote1 for a detailed technical discussion about the email data collection process.)

1

Documentation: Collecting Email for Automated Network Analysis There were three ways that we collected email for network analysis using Condor (available at http://guardian.galaxyadvisors.com/) and to export data to other network analysis programs: (a) Using the Condor network analysis software itself to gather email from a web server, for example, Yahoo or Gmail programs (b) Locating a Microsoft Outlook .pst file on a hard drive and converting it for analysis using Condor (c) Obtaining email in another type of software program, such as Winmail that was built into the Windows Vista operating system 1. Condor and Collecting Email The Condor software program has a mail collector built into it which allowed for retrieving email that is stored on an IMAP or POP server. To retrieve email using Condor, there are basic steps: (a) Open the Fetch Mail function in Condor. (b) Complete the mail form. It is possible to collect just structure (from-to) or both structure and content. (c) Once the “start” is clicked, authentication to the mail server is accomplished, and mailboxes can be selected for download into Condor. It might take a while to download the mail if there are thousands of emails. See the instructional video available on YouTube at: http://youtu.be/ BFo57cz9Jx0.

136

J. C. Gluesing et al.

The commission’s meetings were open to the public, and all the meetings were video recorded. A city staff member took meeting minutes, and they were published on the city’s website. We prepared extended transcripts by watching the videos, and using the meeting minutes as a start, we wrote turn-by-turn notes on what meeting participants said. We then analyzed these extended meeting transcripts qualitatively using the principles of conversation analysis in which each turn is noted according to start and end time for each speaker, and the content of the speaker’s talk is recorded and coded for type of issue frame (IF), for example, regulation, community concern, benefits of drilling, and type of interactional moves or conversational analysis (CA) such as giving information, asking for clarification, and so on (Wasson 2016). To integrate the analysis of the meeting transcripts and the email network data, we took the results of the conversation analysis and reformatted the sequence of speaker turns as network links, as shown in Fig. 2. Each speaker was assigned a unique identification number, A through H. The conversation analysis provided the data to size the nodes by the number of seconds each person spoke during the meeting and to show the link strength between nodes by referencing the number of turns speakers took in conversation with one another. The conversion enabled us to compare the From-To network in a meeting with the From-To email network outside the meeting and apply the structural network centrality measures at the individual and group level to examine interactions and

2. Collecting Email from Outlook on a PC There are multiple versions of Outlook. They all use .pst files. The goal was to archive a .pst of the email and then move it to a USB drive. (a) Create a new .pst (e.g., Name_EAGER.pst). (b) Create a new folder to save the email in the new .pst. (c) Search for emails in Outlook using keywords or just by scrolling through the email box on Outlook. Ideally there is a specific folder (Inbox, Sent, Project), and it is possible to copy the contents of the folder to the newly created .pst. (d) Use Windows Explorer to copy the new .pst to a USB drive. (e) Note: Condor does not analyze any attachments to email, only the email itself. (f) Convert the Outlook .pst emails to a Eudora format that Condor will analyze. For more information about saving data to .pst files, see http://office.microsoft.com/en-us/outlook-help/ create-an-outlook-data-file-pst-to-save-your-information-HA010355677.aspx. 3. Collecting Email from a Winmail Program on a PC Running Vista on Older PCs, Which Is Often the Case in Small Community Governments (a) Open Winmail, and locate the mailbox(es) with the email to be collected for analysis. (b) Create a new folder on the PC’s desktop to store the emails that will be copied for analysis. (c) Scroll through the emails in the Winmail folders, and copy each of the relevant emails into the new desktop folder simply using the “copy” and “paste” commands. The emails will have the extension .eml. Copy the desktop folder to a USB drive. On another PC with Winmail installed and running Microsoft Outlook, copy the folder from the USB drive to the PC’s desktop. Using the import functions, import the emails first into Winmail and then from Winmail into Outlook. Then follow the steps to collect email from Outlook in item number 2 above.

Building Shared Environmental Governance for the Future: The Case of a. . .

137

Fig. 2 Conversion of conversation analysis format to network analysis format

Fig. 3 Example of converting meeting interactions from transcript to network graph

understand the influence of relationships among stakeholders on decisions and how individual social actors might have influenced the decision-making process. Figure 3 is an example of a conversion of meeting interactions into a network graph. The graph is thus a visualization and quantification of the entire meeting. We used the Friedkin and Johnson (2011) social influence network theory and the model for the evolution of social influence networks (Friedkin 2014) to calculate and predict the decision-making, in this case the voting of commission members.

138

J. C. Gluesing et al.

3 Multi-stakeholder Dynamics in Environmental Governance We applied the concept of social influence network theory to the first commission deliberation decision-making meeting and email network data 2 weeks pre- and postmeeting. Figure 4 shows that there were eight people, labeled A through H, who participated in the commission meeting including five commission members, a city facilitator, and two city staff members. The people in the network with the identification numbers P01 through P13 interacted with commission members in the 2 weeks leading up to the first meeting. In the 2 weeks after the meeting, three people joined the email network (labeled A01 through A03) who were not part of the network prior to the meeting. There were also three people (P01, P04, and P05) who participated in the pre-meeting email network who did not take part in the email exchange in the 2 weeks post-meeting. The analysis both inside and outside the meeting led us to several insights about social influence. First, it is clear that the city facilitator, node A, played the most prominent role in the meeting. The facilitator had less of an influence in the interactions pre- and post-meetings, but he was still an important person based upon his eigenvector centrality score. Second, nodes E and F, who represent city staff that interacted with commission members via email outside the meeting, had minimal influence within the meeting interactions. Third, one industry commission member, node B, had consistent influence both outside and inside the meeting. The two members of the commission who represented the voice of the environmentalists, nodes G and H, were also consistent in their influence pre-, during, and post-meeting, but they had less influence in general than the industry members. Fourth, it is possible to see that there was a wider group of people who interacted with the commission members and city staff prior to the first meeting, labeled P01 through P13; however, their influence on commission members appeared to be minimal. This analysis of meetings and email networks over time represented for us a novel methodological approach to investigate multi-stakeholder dynamics in the case of environmental governance.

4 e-Research Approach Our team collaborated online as well as face-to-face to conduct the research we describe in this paper, as the research team was mobile and located in different states across the USA. Our work online fulfilled both social and technical needs. We used an online platform, Amazon Workspaces,2 which allowed us to share our work across both PC and Mac operating systems and provided an automatic backup of our data. However, only one person could access the Workspace at a time, which we

2

http://aws.amazon.com/workspaces/

Building Shared Environmental Governance for the Future: The Case of a. . .

139

Fig. 4 Integration of social influence networks inside the first commission meeting and during the 2 weeks pre- and postmeeting. All nodes are sized by eigenvector centrality. Links are sized by from-to frequency

managed by sending each other texts asking to access the online space and to let each other know when our work was concluded. Any one of us could also demonstrate different techniques and discuss issues such as data preparation and coding and resolve technical problems as they arose. We were able to use a Dropbox folder inside the Amazon Workspace to store and share project documents. Dropbox

140

J. C. Gluesing et al.

facilitated integration of documents created outside the Workspace into our analysis software that we had installed in the Workspace. Online collaboration accommodated our team members’ distributed locations and mobility and reduced our costs because we did not need to purchase dedicated computers; only one software license was required, and the services were free or billed monthly. The study presented in this paper represents an example of what Meyer and Schroeder (2009, 2015) are calling e-Research, “defined as the use of digital tools and data for the distributed and collaborative production of knowledge” (Kindle loc. 91 of 3408). E-Research is based on the understanding that the Internet and its associated infrastructure are enabling collaboration and advances in research practices that cross and connect multiple disciplines and domains (Meyer and Schroeder 2015). The e-Research that supported the collaboration in our own small team has the potential to expand and be extended to a large community of researchers studying environmental governance and governance in general.

5 Conclusion and Opportunities for Future Research In the age of networked governance that crosses both the public and the private sectors and includes multiple stakeholders with differing interests and concerns, it is important to understand how interactions among stakeholders take place over time in multiple ways. Our case example of a community COIN, a commission appointed by a city to help with the creation of a new ordinance to govern oil and gas drilling, illustrates how we can examine both formal and informal interactions. We were able to analyze interactions and social influence among the commission and city officials inside the meetings and pre- and post-meetings among the commissioners and others in the community using publicly available data and the archived emails that commissioners generously granted us permission to use. The methodological approach integrated qualitative and quantitative analysis as well through the conversion of conversation analysis and turn-taking to network data. We suggest that the analysis process illustrated in our example can be scaled to examine networked governance in other multi-stakeholder environments where publicly available data can be obtained at the county, regional, state, or country level. Our analysis was conducted after the ordinance was officially adopted. However, future researchers might attempt to conduct interaction analysis in near real time and perhaps provide feedback and counsel while the decision-making is in progress. With the advances in artificial intelligence and analytical software, it might be possible to automate some of the qualitative data capture and the analysis of qualitative data as well as the data conversion process we describe in our example.

Building Shared Environmental Governance for the Future: The Case of a. . .

141

References Buck, L. E., Geisler, C. C., Schelhas, J., & Wollenberg, E. (Eds.). (2001). Biological diversity: Balancing interests through adaptive collaborative management. Boca Raton: CRC. Carlsson, L., & Sandstrom, A. (2008). Network governance of the commons. International Journal of the Commons, 2(1), 33–54. Dewulf, A., & Bouwen, R. (2012). Issue framing in conversations for change: Discursive interaction strategies for “doing differences”. Journal of Applied Behavioral Science, 48(2), 168–193. Dewulf, A., Gray, B., Putnam, L., & Bouwen, R. (2011). An interactional approach to framing in conflict and negotiation. In W. A. Donohue, R. G. Rogan, & S. Kaufman (Eds.), Framing matters: Perspectives on negotiation research and practice in communication (pp. 7–33). New York: Peter Lang. Folke, C., Hahn, T., Olsson, P., & Norberg, J. (2005). Adaptive governance of social-ecological systems. Annual Review of Environment and Resources, 30, 441–473. Friedkin, N. E. (2014). Social justice in local systems of interpersonal influence. In J. McLeod, E. Lawler, & M. Schwalbe (Eds.), Handbook of the social psychology of inequality (pp. 229–242). New York: Springer. Friedkin, N. E., & Johnson, E. (2011). Social influence network theory: A sociological examination of small group dynamics. Cambridge: Cambridge University Press. Lubell, M. (2013). Governing institutional complexity: The ecology of games framework. Policy Studies Journal, 41(3), 537–559. Meyer, E. T., & Schroeder, R. (2009). The world wide web of research and access to knowledge. Journal of Knowledge Management Research and Practice, 7(3), 218–233. Meyer, E. T., & Schroeder, R. (2015). Knowledge machines: Digital transformations of the sciences and humanities (infrastructures). Cambridge: MIT Press. Nagendra, H., & Ostrom, E. (2012). Polycentric governance of multifunctional forested landscapes. International Journal of the Commons, 6(2), 104–133. Ostrom, E. (2009). A general framework for analyzing sustainability of social-ecological systems. Science, 325(5939), 419–422. Roncoli, C., Orlove, B. S., Kabugo, M. R., & Waiswa, M. M. (2011). Cultural styles of participation in farmers’ discussions of seasonal climate forecasts in Uganda. Agriculture and Human Values, 28(1), 123–138. Wasson, C. (2016). Integrating conversation analysis and issue framing to illuminate collaborative decision-making activities. Discourse & Communication, 10(4), 378–411. Wondolleck, J. M., & Yaffee, S. L. (2000). Making collaboration work: Lessons from innovation in natural resource management. Washington: Island Press.

Effects of Innovation Efficiency and Knowledge on Industry-University Collaboration: An Evolutionary Game Perspective Yang Song and Zhiyuan Zhang

Abstract This paper studies the “evolutionarily stable strategy” (ESS) between industry and university during collaborative innovation processes based on evolutionary games. By designing knowledge sharing models, we analyze the impact factors of knowledge input, knowledge transfer, and innovation cost on collaborative innovation. Furthermore, we use simulation to verify the knowledge sharing model. Our results suggest that the “open innovation strategy” is actually the fact that players choose “evolutionarily stable strategy” in the long-term collaborative innovation process. When the number of game players is different, the small group takes the lead in achieving stabilization strategy. When the number of game players is similar, both groups adopt the “open strategy” at the same speed. Besides, we also suggest that increasing knowledge spillover will contribute to innovation efficiency and stabilization. Theoretically, our study explains the stabilization strategy of the game and provides reasonable recommendations for policy makers.

1 Introduction The collaboration between universities and industries has always been the main concern for innovation in many countries. The existence of university-industry collaboration is of great importance to innovation and economic development (Abramo et al. 2011). The collaboration between industry and universities can not only enhance the innovation ability of firms but also can effectively promote the combination of science, technology, and market potential (Shibayama 2015; Ankrah and AL-Tabbaa 2015). Understanding the impact of research on industrial R&D is central to understanding the innovation process itself (Cohen et al. 2002). Industries are always in need of new technologies due to the competitive marketing needs. Y. Song Jilin University, Changchun, Jilin, China Z. Zhang (*) Jilin University of Finance and Economics, Changchun, China © Springer Nature Switzerland AG 2019 Y. Song et al. (eds.), Collaborative Innovation Networks, Studies on Entrepreneurship, Structural Change and Industrial Dynamics, https://doi.org/10.1007/978-3-030-17238-1_9

143

144

Y. Song and Z. Zhang

Universities are considered as the source of knowledge (Cooke et al. 1997; Conceicao and Heitor 1999; Mazzoleni and Nelson 2005; Asheim et al. 2011; Guerrero et al. 2016), which might have a significant impact upon their local economy (Wright et al. 2008). We cannot ignore the role of universities for regional development or innovation (Cooke et al. 1997; Conceicao and Heitor 1999; Mazzoleni and Nelson 2007; Asheim et al. 2011; Guerrero et al. 2016). The dynamics of the collaboration is influenced by many factors such as upgrading technology (Rosenberg and Nelson 1994; Eichengreen 2004) or knowledge sharing (Mansor et al. 2015; De Silva and Rossi 2018). Previous studies show that industries can get benefits from universities via various ways (Cohen et al. 2002). De Silva and Rossi found that communication capabilities are important for knowledge acquisition between universities and industries (De Silva and Rossi 2018). Collaboration experience, breadth of interaction, and interorganizational trust can lower different types of barriers between universities and industries (Bruneel et al. 2010). Besides, researchers also find numbers of graduate students and numbers of research papers focusing on important factors that promote university-academia collaboration (Yamaguchi et al. 2018). Of all the work mentioned in previous studies, knowledge sharing and resource exchange are still very important for the output and performance of the collaborative alliance (Al-Ashaab et al. 2011; Lee and Ngo 2012). Based on the large interests on university-industry collaboration for both researcher and policy makers, in particular, we will investigate the effects of innovation efficiency and knowledge on industry-university collaboration. This paper is structured as follows: first, we introduce the growing interest of industry-university collaboration in Sect. 1. Second in Sect. 2, we will present our game model between industry and university. In Sect. 3, we will analyze the game model from a theoretical perspective. In the fourth section, we will present the simulating results of innovation efficiency, knowledge complementarity, and knowledge spillover on the collaboration between industry and university. In the last section, we will conclude and discuss the contribution of this research.

2 The Collaborative Innovation Game Model Between Industry and University This paper will analyze the collaborative innovation process between industry and university. In order to understand the innovation process between industry and university, we adopt techniques from evolutionary game theory. Evolutionary game theory has been considered as one of the tools to analyze the collaborative behaviors by researchers and scholars (Nowak and Sigmund 1989; Calvert 1995; Marco and Goetz 2017; Zeng et al. 2017). During the evolution of a game, the players learn from each other, which will influence their future behaviors. Therefore, it is very feasible for us to use evolution game model to study how university and industry can reach a relatively stable collaboration status.

Effects of Innovation Efficiency and Knowledge on Industry-University. . .

2.1

145

Variables and Hypothesis

Currently, due to scientific and technological progress, the competition in the market is increasingly fierce. In order to survive in the market, industries need to maintain a sustainable innovation capability. While technological progress is getting more difficult, complex, and uncertain but stronger integration, industries are unable to cope with various challenges based on independent innovation. Therefore, collaborative innovation between industry and university has been recognized by both policy makers and scholars. Collaborations between industry and university are not only an important means to enhance the innovation ability of industries but also an effective form to promote the combination of science, technology, and economics. Industry and university are regarded as the two main subjects of knowledge creation and spillovers by endogenous growth theory (Romer 1994). However, the industry-university collaboration hasn’t achieved desired results yet because of low transformation rate of scientific research achievement, weak collaboration willingness, unreasonable collaboration model, etc. To a large extent, whether the industry-university collaboration can achieve the desired goal depends on many factors such as the rationality of the players, the collaboration attitude, the expected output of the collaboration, as well as the available resources. These factors lead to the instability of the collaborative innovation process between industry and university. On the one hand, due to the existence of asymmetric information, the “moral hazard” and “adverse selection” between industry and university, players of the collaboration alliance share interests. However, these interests are not exactly the same, which will result in repetitive games between players. On the other hand, the bounded rationality of the participant makes it difficult to find the optimal strategy quickly when facing complex problems. In other words, the process of collaboration is constantly changing and adjusting. The bounded rationality leads to the uncertainty of individual innovative behavior. Therefore, in order to further explain the internal mechanism of industryuniversity collaboration, based on previous research, we propose the following assumptions: Assumption 1 In the economic society there are two agents: industry and university. We indicate the player one from the industry group as E. We indicate player two from the university group as U. Assumption 2 For both industry and universities, there are two strategies for innovation: open innovation and closed innovation. If both players choose the open innovation strategy, they can form a collaborative innovation alliance; if one player chooses closed innovation strategy, they cannot form a collaborative innovation alliance. Assumption 3 The “industry” and “universities” do have a limited rationality due to the information asymmetry. In a sequentially repeated game, they do not necessarily know exactly what strategy the other player will take. But they have some learning capabilities. Each type of agent has its own set of behavioral choices and resulting benefits (payoff structure).

146

Y. Song and Z. Zhang

Assumption 4 The game players are different in terms of innovation capability, knowledge input and output, etc.

2.2

Evolutionary Game Model Construction

In order to design our model, we adopted the model from D’Aspremont and Jacquemin (1988), which suggested that large enough technology spill rate can encourage business collaboration innovation. For the game player one—industry E—and player two, university U, share the same innovation return function π: π ¼ AK α  rK 2

ð1Þ

In this function, A > 0, 0 < a < 1, π is the innovation return. K is knowledge input (such as patent, R&D et al.); A is innovation efficiency, which is the industrial profits formed by knowledge input during innovation process. α is the output elasticity of knowledge input (such as patent conversion rate, etc.). Given that industry and university have different innovation capabilities, they have different innovation input, knowledge input, knowledge output, etc. We specify innovation efficiency of industry as AE. The knowledge input is KE. The output elasticity of knowledge input is αE. The innovation efficiency of university is AU. The knowledge input is KU. The output elasticity of knowledge input is αU. If the industry and the university form a knowledge sharing collaborative innovation alliance, the knowledge input is: ~ ¼ ½ðK E Þρ þ ðK U Þρ 1=ρ K

ð2Þ

ρ is the degree of complementarity of knowledge in the collaborative innovation alliance, and 0 < ρ  1. A lower value of ρ implies a higher degree of complementarity. The existing literature assumes perfect substitutability by setting ρ equal to 1 so that the marginal productivity of R&D investment of each firm is always independent of the investment made by the other firm (Kamien et al. 1992; Anbarci et al. 2002). We define β, β 2 (0,1) as the knowledge spillover coefficient in order to reflect the transfer and absorption of knowledge among players in the collaborative innovation alliance. Based on our Assumption 2, the players have two strategies to choose {open innovation, closed innovation}; there are two types of combinations for the participants’ mode.

2.2.1

The Industry Group and the University Choose Different Innovation Strategy

When “industry” chooses open innovation strategy, “university” chooses closed innovation strategy. The “industry” will invest all the knowledge KE into

Effects of Innovation Efficiency and Knowledge on Industry-University. . .

147

collaborative innovation alliance. The knowledge spillover coefficient is β. The “university” will get the knowledge input from the industry: βKE. As the university chooses closed innovation strategy, the industry won’t get the knowledge input from the university. The knowledge transfer from industry to university is βKE which can lower the research cost of the university to rEUβKE. The additional cost of industry E caused by knowledge spillover is CE. Thus, the payoff function of industry is π 12 , and U the payoff function of the university is π 12 : E ¼ AE K αEE  rK 2E  CE π 12 αU =ρ

U ¼ AU ½βK Eρ þ K Uρ  π 12

 rK 2U þ r EU βK E

ð3Þ ð4Þ

When industry chooses closed innovation strategy, “university” chooses open innovation strategy. The “university” will invest all the knowledge KU into collaborative innovation alliance. The knowledge spillover coefficient is β. The industry will get the knowledge input from the university: βKU. As the “industry” chooses closed innovation strategy, the “university” won’t get the knowledge input from industry. The knowledge transfer from university to industry is βKU which can lower the research cost of the industry to rUEβKU. The additional cost of university caused E by knowledge spillover is CU. Thus, the payoff function of the industry is π 21 , and the U : payoff function of the university is π 21 αE =ρ

E ¼ AE ½K Eρ þ βK Uρ  π 21

 rK 2E þ r UE βK U

U π 21 ¼ AU K αUU  rK 2U  CU

2.2.2

ð5Þ ð6Þ

The Industry Group and the University Group Choose Same Innovation Strategy

When both industry and university choose open innovation strategy, the two players trust each other in the collaborative innovation alliance. The “industry” will invest all the knowledge KE into collaborative innovation alliance. The “university” will invest all the knowledge KU into collaborative innovation alliance. Both sides could obtain all the knowledge input from the other player. The knowledge transfer from university to industry is KU which can lower the research cost of the industry to rUEKU. The knowledge transfer from industry to university is KE which can lower the research cost of the university to rEUKE. Therefore, we have the payoff function E U π 11 for the “industry” and the payoff function π 11 for the “university”: αE =ρ

E ¼ AE ½K Eρ þ K Uρ  π 11

 rK 2E þ r UE K U

ð7Þ

148

Y. Song and Z. Zhang

Table 1 Payoff matrix for industry and university in the innovation alliance Industry group Open innovation: x Closed innovation: 1  x

The university group Open innovation: y

Closed innovation:1  y

E U π 11 ; π 11 E U π 21 ; π 21

E U π 12 ; π 12 E U π 22 ; π 22

αU =ρ

U π 11 ¼ AU ½K Eρ þ K Uρ 

 rK 2U þ r EU K E

ð8Þ

When both industry and university choose closed innovation strategy, in this case, although they may form an innovation alliance, they don’t trust each other. However, they have to invest their knowledge into the innovation alliance to conduct E independent innovation activities. Therefore, we have the payoff function π 22 for the U “industry” and the payoff function π 22 for the “university”: E π 22 ¼ AE K αEE  rK 2E

ð9Þ

U ¼ AU K αUU  rK 2U π 22

ð10Þ

Based on the abovementioned cases, we have the payoff matrices for industry and university of an innovation alliance (Table 1).

3 Construction and Analysis of Game Model 3.1

Solution to the Game Model

Assume at time t, if the ratio of open innovation strategy of the industry is x(t), x (t) 2 [0, 1], then the ratio of closed innovation strategy of the industry is 1  x(t). If the ratio of open innovation strategy of the university is y(t), y(t) 2 [0, 1], then the ration of closed innovation strategy of the university is 1  y(t). Given the current period x(t), the next phase to take open innovation and closed innovation strategies for the industry are: uH ¼ yðt Þπ 111 þ ð1  yðt ÞÞπ 112

ð11Þ

uB ¼ yðt Þπ 121 þ ð1  yðt ÞÞπ 122

ð12Þ

Therefore, the average payoff of the industry group is: u ¼ xðt ÞuH þ ð1  xðt ÞÞuB

ð13Þ

We define the current period as y(t). When adopting open innovation strategy and closed innovation strategy, the expected payoff of the next period for the university is:

Effects of Innovation Efficiency and Knowledge on Industry-University. . .

149

vH ¼ xðt Þπ 211 þ ð1  xðt ÞÞπ 221

ð14Þ

vB ¼ xðt Þπ 212 þ ð1  xðt ÞÞπ 222

ð15Þ

Therefore, the average payoff for the university is: v ¼ yðt ÞvH þ ð1  yðt ÞÞvB

ð16Þ

When mentioning the dynamic change speed of x(t) and y(t), we adopted the dynamic equations from previous study (Amann and Possajennikov 2009):


E E E E F ðxÞ ¼ dx=dt ¼ xð1  xÞ y π 11 þ ð1  yÞ π 12  π 21  π 22 We define
 F(x) ¼
0 and
get x1 ¼ 0,  E E E E E E y ¼ π 12  π 22 = π 12  π 22  π 11  π 21 Similarly, we have the dynamic equation for university group: 

x2 ¼ 1,




U U U U F ðyÞ ¼ dy=dt ¼ yð1  yÞ x π 11  π 12  π 22 þ ð1  xÞ π 21

ð17Þ and

ð18Þ

y2 ¼ 1, and We define
 F( y) ¼
0 and
get y1 ¼ 0,  U U U U U U = π 21  π 11  π 22  π 22  π 12 x ¼ π 21 We combine Eqs. (17) and (18) and take industry and university as a system; the system has five special dynamic system balance points, which are (0,0), (0,1), (1,0), (1,1), and (x,y). According to the local stability of the Jacobi matrix, the local stability of the system in these dynamic systems is analyzed. Using the Eqs. (17) and (18), we can calculate the sign and trace of the matrix determinant at the equilibrium point (x ¼ 0, y ¼ 0): The determinant of J is:
U
E E U π 12  π 22 π 21  π 22 >0

ð19Þ



E E U U  π 22  π 22 π 12 þ π 21 SMZGH, the probabilities which industry and university evolve to the direction of choosing open innovation strategy are bigger than choosing closed innovation strategy. If SOZGH ¼ SMZGH, the probabilities which industry and university evolve to the direction of choosing open innovation strategy are equal to the probabilities of choosing closed innovation strategy. Therefore, according to Fig. 1, the relationship between the size of SOZGH and SMZGH area directly determines the probability that the industry chooses open innovation strategy and the probability that the university chooses true disclosure strategy. As SOZGH + SMZGH ¼ 1, we only need to discuss the size of SOZGH. The area of SOZGH can be decomposed into the area of triangle area SOZG and the area of triangle area SOHG. Therefore, the factors influencing the probabilities of both industry and university choosing open innovation strategy can be analyzed by the factors influencing the size of area SOZG and area SOHG. Similarly, the size of area SMZGH determines the probability that industry chooses open innovation strategy. As SOLGN + SMLGN ¼ 1, the influence of the same factor on the probability of the open innovation strategy chosen by industry and the probability of the open innovation strategy chosen by university has an inverse relationship. When analyzing the factors that affect the probability of the industry choosing open innovation strategy and the factors that affect the size of the university choosing open innovation strategy, we use our assumptions.
 U

 U U U U U SOHG ¼ x1 =2 ¼ π 21 = π 21  π 22  π 11 =2  π 22  π 12


 E E E E E E = π 12  π 11 =2 SOZG ¼ y2 =2 ¼ π 12  π 22  π 22  π 21

ð21Þ ð22Þ

In order to intuitively see the effects of the parameters contained in Eqs. (21) and (22) on the path of evolution of strategy choices of both players, specifically, the influence of factors such as the amount of knowledge input, the level of knowledge spillover, and the cost of innovation on the evolution path of industry and university strategy choices, Matlab 2014a software will be used as a numerical simulation tool to simulate the evolution path of strategy choice of both players.

152

Y. Song and Z. Zhang

4 Numerical Simulation We use MATLAB 2014a for the numerical simulation. We analyze factors that affect the evolution path in the collaborative innovation alliance. We also analyze the impact factors on the stability of collaborative innovation alliance. The numerical simulation can provide the insights about how the variables influence the evolution path and give extra evidence to the game model.

4.1

The Impact of Innovation Efficiency A

Given that the industry and the university have different innovation capabilities, they have different knowledge input and knowledge output. We assume the innovation efficiency of the industry AE ¼ 0.5, the knowledge input KE ¼ 3, the output elasticity of knowledge input αE ¼ 1, the knowledge spillover coefficient β ¼ 0.5, the additional cost CE ¼ 2, the innovation efficiency of the university AU ¼ 0.5, the knowledge input KU ¼ 2, the output elasticity of knowledge input αU ¼ 1, the additional cost CU ¼ 2, the degree of complementarity of knowledge ρ ¼ 1, the knowledge input cost coefficient r ¼ 0.5, the industry innovation cost reduction factor rUE ¼ 0.2, and the university innovation cost reduction factor rEU ¼ 0.2. Now we increase the innovation efficiency of industry from AE ¼ 0.5 to AE ¼ 1 and get Fig. 2. We increase the innovation efficiency of university from AU ¼ 0.5 to AU ¼ 1 and get Fig. 3. As shown in Fig. 2, the black solid evolution curve is the evolution path of the industry when innovation efficiency AE ¼ 0.5. The five curves represent the probability that industry chooses open innovation strategy from x(t) ¼ 0.1 to x(t) ¼ 0.9, respectively. As for industry group, because of the low innovation efficiency, regardless of the probability of initial choice of open innovation strategy in the industry group, the final industry group is evolving toward to the choice of closed innovation strategy. The red dotted evolution curve is the evolution path of the industry when innovation efficiency AE ¼ 1. The five curves represent the probability of choosing open innovation strategy x(t) ¼ 0.1 to x(t) ¼ 0.9, respectively. As for the industry group, because of the improvement of innovation efficiency, regardless of the probability of initial choice of open innovation strategy in the industry group, the final industry group is evolving toward the choice of close innovation strategy. However, the evolutionary speed is obviously getting slower. Therefore, no matter what the probability of choosing open innovation strategy is, the industry group will evolve toward to closed innovation strategy. As shown in Fig. 3, the black solid evolution curve is the evolution path of the university when innovation efficiency AU ¼ 0.5. The five curves represent the probability that university chooses open innovation strategy from y(t) ¼ 0.1 to y (t) ¼ 0.9, respectively. As for university group, because of the low innovation efficiency, regardless of the probability of initial choice of open innovation strategy, the final university group is evolving toward the choice of closed innovation

Effects of Innovation Efficiency and Knowledge on Industry-University. . .

153

The probability of enterprise strategy x

0.9

Innovation efficiency AE =0.5 Innovation efficiency AE =1 0.7

0.5

0.3

0.1 0

2

4

6

time t

8

10

12

Fig. 2 Increasing innovation efficiency on industry’s evolutionary path

strategy. The red dotted evolution curve is the evolution path of the university when innovation efficiency AE ¼ 1. The five curves represent the probability that university chooses open innovation strategy from y(t) ¼ 0.1 to y(t) ¼ 0.9, respectively. As for the university group, because of the improvement of innovation efficiency, regardless of the probability of the initial choice of open innovation strategy in the industry group, the final industry group is evolving toward to the choice of an open innovation strategy. Through the comparative analysis of Figs. 2 and 3, when improving the efficiency of innovation, the industry group will still choose closed innovation strategy at a relatively slower evolving speed. However, the university will actively choose an open innovation strategy. Meanwhile, the speed of evolving toward choosing open innovation strategy is significantly faster than evolving toward choosing closed innovation strategy. As shown in Fig. 2, when the innovation efficiency is relatively lower, when t ¼ 3, the probability of choosing a closed innovation strategy for the industry is almost stable to x(t) ¼ 0. While when the innovation efficiency is relatively higher, when t ¼ 10, the probability of choosing a closed innovation strategy for the industry is almost stable to x(t) ¼ 0. As shown in Fig. 3, when the innovation efficiency is relatively lower, after t ¼ 10, the probability of choosing a closed innovation strategy for the university is almost stable to y(t) ¼ 0. While when the innovation efficiency is relatively higher,

154

Y. Song and Z. Zhang

The probability of University strategy y

0.9

0.7

Innovation efficiency AE =0.5 Innovation efficiency AE =1 0.5

0.3

0.1 0

2

4

6

time t

8

10

12

Fig. 3 Increasing innovation efficiency on university’s evolutionary path

when t ¼ 8, the probability of choosing an open innovation strategy for the university is almost stable to y(t) ¼ 1. In other words, the university group can evolve and stabilize faster to the equilibrium point than the industry group. In addition, the university group is more likely to choose an open innovation strategy. This is because the knowledge input and the cost of the industry is higher than the knowledge input of university. Therefore, as for the collaborative innovation alliance, the game side which invest lower knowledge input and cost is more willing to form the collaborative innovation alliance.

4.2

The Impact of the Degree of Complementarity of Knowledge ρ on Collaborative Innovation Alliance

We keep most of the simulation parameters in Sect. 4.1. We assume the innovation efficiency of industry AE ¼ 0.5, the knowledge input KE ¼ 3, the output elasticity of knowledge input αE ¼ 1, the additional cost CE ¼ 2, the innovation efficiency of university AU ¼ 0.5, the knowledge input KU ¼ 2, the output elasticity of knowledge input αU ¼ 1, the additional cost CU ¼ 2, and the knowledge spillover coefficient β ¼ 0.5. Now we change the degree of complementarity of knowledge from ρ ¼ 1 to

Effects of Innovation Efficiency and Knowledge on Industry-University. . .

155

The probability of enterprise strategy x

0.9

0.7

0.5

0.3

Knowledge Complementarity =1 0.1 0

Knowledge Complementarity =0.5 2

4

6

time t

8

10

12

Fig. 4 Complementarity of knowledge on industry’s evolutionary path

ρ ¼ 0.5; we get Figs. 4 and 5 to observe the evolutionary path of player one industry and player two university. As shown in Fig. 4, the black solid evolution curve is the evolution path of the industry when the degree of complementarity of knowledge ρ ¼ 1. The five curves represent the probability that industry chooses an open innovation strategy which is from x(t) ¼ 0.1 to x(t) ¼ 0.9, respectively. As for the industry group, because of the higher degree of complementarity of knowledge, regardless of the probability of an initial choice of open innovation strategy in the industry group, the final industry group is evolving toward to the choice of a closed innovation strategy. The red dotted evolution curve is the evolution path of the industry when the degree of complementarity of knowledge ρ ¼ 0.5. The five curves represent the probability of choosing open innovation strategy which is from x(t) ¼ 0.1 to x(t) ¼ 0.9, respectively. As for the industry group, because of the lower degree of complementarity of knowledge, regardless of the probability of an initial choice of open innovation strategy in the industry group, the final industry group is evolving toward the choice of an open innovation strategy. As shown in Fig. 5, the black solid evolution curve is the evolution path of the university when the degree of complementarity of knowledge ρ ¼ 1. The five curves represent the probability that the university chooses an open innovation strategy

156

Y. Song and Z. Zhang 1

The probability of University strategy y

0.9

Knowledge Complementarity =1

0.8

Knowledge Complementarity =0.5 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

0

2

4

6

time t

8

10

12

Fig. 5 Complementarity of knowledge on university’s evolutionary path

which is from y(t) ¼ 0.1 to y(t) ¼ 0.9, respectively. As for the university group, because of the higher degree of complementarity of knowledge, regardless of the probability of initial choice of an open innovation strategy in the university group, the final university group is evolving toward to the choice of a closed innovation strategy. The red dotted evolution curve is the evolution path of the university when degree of complementarity of knowledge ρ ¼ 0.5. The five curves represent the probability of choosing an open innovation strategy which is from y(t) ¼ 0.1 to y (t) ¼ 0.9, respectively. As for the university group, because of the lower degree of complementarity of knowledge, regardless of the probability of initial choice of an open innovation strategy in the university group, the final university group is evolving toward the choice of an open innovation strategy. Through the comparative analysis of Figs. 4 and 5, whether it is an industry group or university group, when the degree of complementarity of knowledge is lower, both game sides will actively choose an open innovation strategy. When the degree of complementarity of knowledge is higher, the speed that the university evolves toward to choosing a closed innovation strategy is significantly slower than that for the industry group. As shown in Fig. 4, when t ¼ 3, the probability of choosing an open innovation strategy for the industry is almost stable to x(t) ¼ 0. As shown in Fig. 5, after t ¼ 12, the probability that the university choose a closed innovation strategy is nearly stable to y(t) ¼ 0. When the degree of complementarity of

Effects of Innovation Efficiency and Knowledge on Industry-University. . .

157

knowledge is relatively lower, the speed that the university evolves toward to choosing an open innovation strategy is significantly faster than that for the industry group. As shown in Fig. 4, when t ¼ 12, the probability of choosing an open innovation strategy for the industry is nearly stable to x(t) ¼ 1. As shown in Fig. 5, when t ¼ 6, the probability of choosing an open innovation strategy for the university is nearly stable to y(t) ¼ 1. This is because the industry group invests a large amount of knowledge and spends more on the cost, while the knowledge spillover has less effect on the cost reduction. Therefore, for collaborative innovation alliances, the less input knowledge, the more obvious the effect of knowledge spillover on cost reduction and the more it hopes to form a synergistic innovation alliance as soon as possible. In addition, comparing Figs. 2 and 4, the evolution of the industry groups to open innovation strategy is faster. Comparing Figs. 3 and 4, universities and research institutes have evolved faster toward cooperative innovation strategies. This shows that reducing the complementarity of knowledge is more effective than providing innovation efficiency to the formation of collaborative innovation alliances.

4.3

The Knowledge Spillover Coefficient β on Collaborative Innovation Alliance

We keep most of the simulation parameters in Sect. 4.2. We assume the innovation efficiency of industry AE ¼ 0.5, the knowledge input KE ¼ 3, the output elasticity of knowledge input αE ¼ 1, the additional cost CE ¼ 2, the innovation efficiency of university AU ¼ 0.5, the knowledge input KU ¼ 2, the output elasticity of knowledge input αU ¼ 1, the additional cost CU ¼ 2, and the degree of complementarity of knowledge ρ ¼ 1. Since we already have the figure when β ¼ 0.5 in Sect. 4.2, both game sides will choose open innovation strategy to form the collaborative innovation alliance. Now we adjust the knowledge spillover coefficient from β ¼ 0.5 to β ¼ 0.7 then to β ¼ 1. We get the Figs. 6 and 7 to observe the evolutionary path of player one industry and player two university. As shown in Fig. 6, the red dotted evolution curve is the evolution path of the industry when the knowledge spillover coefficient β ¼ 0.5. The black solid evolution curve is the evolution path of the industry when the knowledge spillover coefficient β ¼ 0.7. The blue dotted evolution curve is the evolution path of the industry when the knowledge spillover coefficient β ¼ 1. All the three evolution curves start from a point where the probability of open innovation x(t) ¼ 0.5. As for the industry group, because the knowledge spillover coefficient β ¼ 0.5 is a lower value, regardless of the probability of initial choice of an open innovation strategy in the industry group, the final industry group is evolving toward the choice of an open innovation strategy. When the knowledge spillover coefficient is improved to β ¼ 0.7 and β ¼ 1, because the knowledge spillover coefficient is a higher value, regardless of the probability of an initial choice of an open innovation strategy in the industry group, the final

158

Y. Song and Z. Zhang

The probability of enterprise strategy x

0.9

0.7

Knowledge spillover coefficient =0.5 Knowledge spillover coefficient =0.7

0.5

Knowledge spillover coefficient =1

0.3

0.1 0

2

4

6

time t

8

10

12

Fig. 6 Knowledge spillover on industry’s evolutionary path

industry group is evolving toward the choice of a closed innovation strategy. Through the comparison, we can conclude that a higher level of knowledge spillovers can lead to free-rider problems from both sides of the game. Therefore, the game strategy of the industry will evolve to closed innovation strategy. As shown in Fig. 7, the red dotted evolution curve is the evolution path of the university when the knowledge spillover coefficient β ¼ 0.5. The black solid evolution curve is the evolution path of the university when the knowledge spillover coefficient β ¼ 0.7. The blue dotted evolution curve is the evolution path of the university when the knowledge spillover coefficient β ¼ 1. All the three evolution curves start from a point where the probability of open innovation y(t) ¼ 0.5. As for the university group, because the knowledge spillover coefficient β ¼ 0.5 is a lower value, regardless of the probability of an initial choice of an open innovation strategy in the university group, the final university group is evolving toward the choice of an open innovation strategy. When the knowledge spillover coefficient is improved to β ¼ 0.7 and β ¼ 1, because the knowledge spillover coefficient is a higher value, regardless of the probability of an initial choice of an open innovation strategy in the university group, the final university group is evolving toward the choice of a closed innovation strategy. Through the comparison, we can conclude that a higher level of knowledge spillovers can lead to free-rider problems from both sides of the game. Therefore, the game strategy of the university will evolve to a closed innovation strategy.

Effects of Innovation Efficiency and Knowledge on Industry-University. . .

159

The probability of University strategy y

0.9

0.7

0.5

Knowledge spillover coefficient =0.5 Knowledge spillover coefficient =0.7

0.3

Knowledge spillover coefficient =1 0.1 0

2

4

6

time t

8

10

12

Fig. 7 Knowledge spillover on university’s evolutionary path

5 Conclusion Collaborative innovation is the innovation behaviors of multiple subjects. The inherent mechanism is to discuss the possibility of collaborative innovation under uncertain conditions and to achieve the stability of collaboration and institutionalization. Based on evolutionary game theory, this paper analyzes the repeated game mechanism of industry, university from the perspective of bounded rationality. We find a few major points based on our model construction and simulation. First of all, in the long-term collaborative innovation, if the innovation efficiency is high enough, industries, universities, and research institute will eventually collaborate to achieve the “evolution stabilization strategy” of collaborative innovation. Secondly, if we lower the degree of complementarity of knowledge, both industry and university will choose an open innovation and form a collaborative innovation alliance. Meanwhile, the player who has higher knowledge input is more willing to choose an open innovation strategy. Compared to innovation efficiency, knowledge complementarity is more effective for collaborative innovation alliance. Last but not least, the most effective way to solve a free ride in a collaborative innovation alliance is to lower the possibility of knowledge spillover. When the level of knowledge spillover is too high, the game players as “economic man” will choose

160

Y. Song and Z. Zhang

a closed innovation strategy rather than an open innovation strategy in order to improve the innovation output. Based on abovementioned conclusions, the practical significance is that industries can increase collaboration with the university to improve their own innovation capacity in developed industry areas. A university should work on the topics which are relevant to regional development in order to improve research efficiency. The government should make use of policy guidance, developing intermediary organizations, and system construction to increase the trust between the industry and the university. This will contribute to the efficiency of collaborative innovation. One of the limitations in our paper is that we only considered the effect of knowledge input on innovation return, which constrained our theoretical analysis. Given that there are more factors such as innovation cost, default penalties, the game model should be further developed in the next step.

References Abramo, G., D’Angelo, C. A., & Di Costa, F. (2011). University-industry research collaboration: A model to assess university capability. Higher Education, 62, 163–181. https://doi.org/10.1007/ s10734-010-9372-0 Al-Ashaab, A., Flores, M., Doultsinou, A., & Magyar, A. (2011). A balanced scorecard for measuring the impact of industry–university collaboration. Production Planning and Control, 22, 554–570. https://doi.org/10.1080/09537287.2010.536626 Amann, E., & Possajennikov, A. (2009). On the stability of evolutionary dynamics in games with incomplete information. Mathematical Social Sciences, 58, 310–321. https://doi.org/10.1016/J. MATHSOCSCI.2009.08.001 Anbarci, N., Lemke, R., & Roy, S. (2002). Inter-firm complementarities in R&D: A re-examination of the relative performance of joint ventures. International Journal of Industrial Organization, 20, 191–213. https://doi.org/10.1016/S0167-7187(00)00081-3 Ankrah, S., & AL-Tabbaa, O. (2015). Universities–industry collaboration: A systematic review. Scandinavian Journal of Management, 31, 387–408. https://doi.org/10.1016/J.SCAMAN.2015. 02.003 Asheim, B. T., Smith, H. L., & Oughton, C. (2011). Regional innovation systems: Theory, empirics and policy. Regional Studies, 45, 875–891. https://doi.org/10.1080/00343404.2011.596701 Bruneel, J., D’Este, P., & Salter, A. (2010). Investigating the factors that diminish the barriers to university–industry collaboration. Research Policy, 39, 858–868. https://doi.org/10.1016/J. RESPOL.2010.03.006 Calvert, R. L. (1995). The rational choice theory of institutions: Implications for design. In Institutional design (pp. 63–94). Dordrecht: Springer Netherlands. Cohen, W. M., Nelson, R. R., & Walsh, J. P. (2002). Links and impacts: The influence of public research on industrial R&D. Management Science, 48, 1–23. https://doi.org/10.1287/mnsc.48.1. 1.14273 Conceicao, P., & Heitor, M. V. (1999). On the role of the university in the knowledge economy. Science and Public Policy, 26, 37–51. https://doi.org/10.3152/147154399781782617 Cooke, P., Gomez Uranga, M., & Etxebarria, G. (1997). Regional innovation systems: Institutional and organisational dimensions. Research Policy, 26, 475–491. https://doi.org/10.1016/S00487333(97)00025-5

Effects of Innovation Efficiency and Knowledge on Industry-University. . .

161

D’Aspremont, C., & Jacquemin, A. (1988). Cooperative and noncooperative R & D in duopoly with spillovers. The American Economic Review, 78, 1133–1137. https://doi.org/10.1016/0014-2921 (88)90202-4 De Silva, M., & Rossi, F. (2018). The effect of firms’ relational capabilities on knowledge acquisition and co-creation with universities. Technological Forecasting and Social Change, 133, 72–84. https://doi.org/10.1016/J.TECHFORE.2018.03.004 Eichengreen, B. (2004). Productivity growth, the new economy, and catching up. Review of International Economics, 12, 243–245. https://doi.org/10.1111/j.1467-9396.2004.00446.x Guerrero, M., Urbano, D., Fayolle, A., et al. (2016). Entrepreneurial universities: Emerging models in the new social and economic landscape. Small Business Economics, 47, 551–563. https://doi. org/10.1007/s11187-016-9755-4 Kamien, M., Muller, E., & Zang, I. (1992). Research joint ventures and R&D cartels. The American Economic Review, 82, 1293–1306. Lee, S., & Ngo, T. (2012). The capitalization of knowledge: A triple helix of university-industrygovernment. Higher Education, 63, 161–163. Mansor, Z. D., Mustaffa, M., & Salleh, L. M. (2015). Motivation and willingness to participate in knowledge sharing activities among academics in a public university. Procedia Economics and Finance, 31, 286–293. https://doi.org/10.1016/S2212-5671(15)01188-0 Marco, J., & Goetz, R.-U. (2017). Tragedy of the commons and evolutionary games in social networks: The economics of social punishment. FEEM Working Papers. https://doi.org/10. 2139/ssrn.2998546 Mazzoleni, R., & Nelson, R. R. (2005). The roles of research at universities and public labs in economic catch-up (No. 2006/01). LEM Working Paper Series. Mazzoleni, R., & Nelson, R. R. (2007). Public research institutions and economic catch-up. Research Policy, 36, 1512–1528. https://doi.org/10.1016/j.respol.2007.06.007 Nowak, M., & Sigmund, K. (1989). Game-dynamical aspects of the prisoner’s dilemma. Applied Mathematics and Computation, 30, 191–213. https://doi.org/10.1016/0096-3003(89)90052-0 Pardalos, P. M., Migdalas, A., & Pitsoulis, L. (2008). Pareto optimality, game theory and equilibria. New York: Springer. Romer, P. M. (1994). The origins of endogenous growth. The Journal of Economic Perspectives, 8, 3–22. https://doi.org/10.1257/jep.8.1.3 Rosenberg, N., & Nelson, R. R. (1994). American universities and technical advance in industry. Research Policy, 23, 323–348. https://doi.org/10.1016/0048-7333(94)90042-6 Shibayama, S. (2015). Academic commercialization and changing nature of academic cooperation. Journal of Evolutionary Economics, 25, 513–532. https://doi.org/10.1007/s00191-014-0387-z Wright, M., Clarysse, B., Lockett, A., & Knockaert, M. (2008). Mid-range universities’ linkages with industry: Knowledge types and the role of intermediaries. Research Policy, 37, 1205–1223. https://doi.org/10.1016/J.RESPOL.2008.04.021 Yamaguchi, Y., Fujimoto, J., Yamazaki, A., & Koshiyama, T. (2018). A study of the factors influencing industry-academia collaboration activities in private universities. In 2018 Portland International Conference on Management of Engineering and Technology (PICMET) (pp. 1–10). Honolulu, HI: IEEE. Zeng, W., Li, M., & Feng, N. (2017). The effects of heterogeneous interaction and risk attitude adaptation on the evolution of cooperation. Journal of Evolutionary Economics, 27, 435–459. https://doi.org/10.1007/s00191-016-0489-x

Part IV

Creativity

Measuring Human-Animal Interaction with Smartwatches: An Initial Experiment Katharina Stolz, Teresa Heyder, Peter A. Gloor, and Oliver Posegga

Abstract The paper describes and evaluates an explorative approach to quantify the relationship between trainable animals and their owners. Data on human-animal interaction has been collected by using Pebble smartwatches and by observing different kinds of animal training sessions. Tracking movement of horses and dogs with the Pebble Watch was successful with horses but not with dogs. Besides the breed and behavior of the animal, weather conditions and the way of attaching the Pebble influenced the measurement quality. In summary, the experiment indicates that there might be a connection between the heart rate (BPM), the average movement (VMC), and the mood data (pleasance and activation) of an animal and its owner during training sessions.

1 Introduction In the last years, there has been a rise in the popularity of fitness trackers worldwide. Compared to 2016 with a sales revenue of about 2.49 billion US dollars, it increased to around 2.51 billion US dollars in 2017 and is forecasted to 3.33 billion US dollars by 2022 (Statista 2018). With their growing popularity, a plethora of self-tracking technologies has become available to consumers. The Quantified Self Labs (2011) homepage lists 505 self-tracking tools, including wearable devices, websites, and mobile applications. Those tools enable individuals to track and analyze their athletic performance and health issues (Budzinski and Schneider 2017). Besides measuring body signals of humans, fitness trackers and smartwatches might also be used to

K. Stolz (*) · T. Heyder · O. Posegga Otto-Friedrich-University of Bamberg, Bamberg, Germany e-mail: [email protected]; [email protected]; [email protected] P. A. Gloor MIT Center for Collective Intelligence, Cambridge, MA, USA e-mail: [email protected] © Springer Nature Switzerland AG 2019 Y. Song et al. (eds.), Collaborative Innovation Networks, Studies on Entrepreneurship, Structural Change and Industrial Dynamics, https://doi.org/10.1007/978-3-030-17238-1_10

165

166

K. Stolz et al.

measure those of trainable animals, like dogs and horses (e.g., US 6,311,645 B1, 2001). During a training session with an animal, e.g., a horse, the measurement of the heart rate allows the trainer to monitor basic vital values and detect signs of tiredness, sickness, and injury in real time (Polar Electro 2017b). Based on insights gained through tracking technologies, training sessions can be designed more efficiently. Further, they enable the trainer to better understand the animal and its reactions, which helps to avoid under- and overtraining and to realize the animal’s full potential (Evans 2000). Hence, the question arises, whether more information can be derived by connecting the results of the animal’s and owner’s self-tracking tool, e.g., impacts on the well-being of both parties during and after training sessions. This paper explores the potential of self-tracking devices in measuring human-animal interaction (HAI) by answering the following questions: First, can the Pebble Watch be used as a tracking tool for biometric signals of animals? Second, what are the factors influencing the measurement quality? And third, can combined tracking of animal and owner signals be used to gain insights into the connection between animal and owner? Thereby, the focus lays on horses and dogs as trainable (e.g., US 6,311,645 B1, 2001) and highly social animals (e.g., Deine Tierwelt GmbH and Co. KG 2015; Evans 2000). The paper starts with an overview of the current state of the art in HAI research and types of training with dogs and horses in Sect. 2. In Sect. 3, the paper examines how to measure HAI using affordable, readily available, and mobile tracking technologies. Section 4 presents an overview of the methods used in this study and introduces the Pebble Watch as the primary tracking technology used in this study. Section 5 contains the findings and their discussion. The paper closes with limitations and potential avenues for future research.

2 Related Work HAI has gained a considerable amount of scholarly attention during the past years. The literature review by Beetz et al. (2012) discusses 69 HAI studies and serves as the foundation for this section. In the following, we summarize some of their central findings, whereinafter basic types of training sessions of dogs and horses influencing the HAI are described: Beetz et al. (2012) identify eight studies, which report a reduced heart rate while or after the interaction between humans and animals. Allen et al. (1991) measured the “autonomic reactivity,” including the pulse rate, skin conductance, and blood pressure, of 45 adult women performing a stress task in a laboratory and 2 weeks later in their home while either being accompanied by a friend or their dog or being alone. The results show that the participants exhibit little or even no physiological reactivity while performing a stressful task when their dog was present and a greater reactivity when their closest friend was present (Allen et al. 1991). According to Gehrke (2010), negative emotions like “sadness, fear, anger, or frustration” lead to

Measuring Human-Animal Interaction with Smartwatches: An Initial Experiment

167

more erratic and imbalanced heart rhythms, whereas positive emotions like “joy, appreciation, and love” lead to a more ordered and balanced heart rhythm as a result of “autonomic nervous system responses.” Further, Gehrke (2010) observed that the HRV frequency dynamics of humans and horses are similar. The investigation indicates that in six of the seven human-horse pairs, persons involved synchronized their “HRV frequency cycle” to match the one of the horse but not vice versa (Gehrke 2010). According to these results, previous research indicates a relationship between the animal’s HRV frequency and those of the owner. Four studies of the literature review report a positive influence of HAI on the (perceived) human stress level. Barker et al. (2005) compare the stress levels of individuals who either rest for 20 min or pet an animal for 5–20 min. They conclude that the reduction of stress is higher when petting a dog. According to Allen et al. (2002), pets can reduce the reaction to acute stress and also decrease the perceived stress. In addition, pet owners, compared to nonowners, recover faster from elevated stress levels. The initial findings by Gehrke (2010) also suggest that when the horse is relaxed and the human being feels stressed, the horse might “transfer” the calmness to the human so that the individual feels more relaxed. Before this background, the experiment conducted in this study explores the effect of HAI on the owner’s self-reported stress. Wells (2009) concludes that although the evidence for a direct causal relationship between animals and human well-being is not yet conclusive, the literature widely supports that pets are good for human beings. Hence, the owner might experience a higher self-reported pleasance at the end of an HAI than before. Therefore, the HAI might impinge on the welfare of the owner. Regarding Góngora and Solano (2015), the pleasant life with α ¼ 0.81 (Cronbach’s Alpha) is an adequate indicator of measuring the well-being. This present experiment will explore possible impacts of the HAI on the pleasance of the owner as one indicator for well-being. Summarizing the above, the aim of the study is to measure the animal’s and human’s heart rate and movement as well as the owner’s activation, pleasance, and his/her self-reported basic emotions (including stress) by interacting with trainable animals over the course of time. As the study focusses on HAI with dogs and horses as trainable animals that live in packs, the diverse types of training (nonprofessional) with separate aims and consequences will be considered hereafter. First, groundwork (Pferdeflüsterei 2017b) is the basic prerequisite of dealing with horses successfully and comprises training via a direct connection with the horse through a lunge, rope, or halter without being on the saddle. Basic skills like hand changing of the horse, standing still during cleaning, giving hoofs, and stopping or moving on command are trained. Groundwork even starts with little feedback to the horse in the stable. Threatening gestures, e.g., laid-back ears or bare teeth, show that the horse does not accept its owner vs. gestures of friendship like rumbling or nudging with their nose. Analogously, the basic commands are important for the human-dog relation wherefore at least the commands “sit,” “down,” and “wait” should be trained (Deine Tierwelt GmbH and Co. KG 2015). Second, the round pen (Pferdeweb 2017) is a round and fenced place, where the horse is walking around its owner without a physical

168

K. Stolz et al.

connection between both (e.g., neither lunge, rope, or halter). The owner directs his/her animal only via body language. It is usually used for dominance practice of western horses and is only ending when the horse accepts the owner as its lead animal. By moving the head around, rearing up or rejecting, the horse is trying to test its limits. This can be compared with dominance training by dogs, where the dog accepts its owner as the “pack leader” (Deine Tierwelt GmbH and Co. KG 2015). Third, riding (Pferdeflüsterei 2017a) on a saddle, the muscles, the sense of balance, the condition, and the dexterity of the horses are trained. The focus lies on optimizing the skills by overcoming obstacles and subtle changes from gait to gait. It is important that the owner is able to parry the horse. If the horse does not accept the rider as the leading animal, the horse won’t listen and can buck the rider off or refuse barriers. Groundwork or round pen sessions are building the foundation for riding. A typical riding session in this experiment takes about 1 h and includes warm-up and cooldown phases. When the horse is snorting in relaxation phases, it can be seen as a sign of satisfaction, whereas squeaking or a re-deemed tail indicates fear or pain (exercise should be stopped). Taking a dog for a walk could be compared to riding, as the basic commands can be seen as a prerequisite of a stressless walk (Deine Tierwelt GmbH and Co. KG 2015).

3 Method For our research, we used the Pebble Watch 2 that was designed to measure, e.g., human heart rate (BPM) and the “vector magnitude counts” (VMC) to improve the owner’s performance and to achieve fitness goals (Pebble; Pebble 2016). To measure the heart rate, the watch uses “optical heart rate monitors” by sending hundreds of flashes each second (Pebble Support 2016). The heart rate monitor technology measures the beats per minute on a periodic basis, whereby the frequency of measurement “varies and depends on the level and activity of the user” (Pebble 2016). VMC is “a measure of the total amount of movement seen by the watch” (Pebble). To attach the watch to the animal, lashing straps were used (see Fig. 1) as it is possible to adjust the size of the strap according to the thighbone/neck circumference of the animal (close to the main artery) and as it is enough flexible, wherefore the lasing strap causes no injuries. Thereby the heart rate (BPM) and movement (VMC) of the animal were measured. In addition, the Happimeter extension (Budner et al. 2017) was used to collect data on pleasance and activation of the owner. These two variables are based on the Circumplex Model of affect (Posner et al. 2005) and are built on a nine-outcome grid. Thereby the variables are defined as linear, with a scale from 1 to 3, representing the dimensions high pleasance/activation, medium pleasance/activation, and low pleasance/activation (Budner et al. 2017). Data is gathered by either entering the pleasance and activation manually through self-reporting or a learning algorithm that simultaneously records the dimensions of pleasance and activation based on the

Measuring Human-Animal Interaction with Smartwatches: An Initial Experiment

169

Fig. 1 Attachment of the Pebble Watch

prediction model including several variables (e.g., temperature, windiness, air pressure, GPS, VMC, activity, and heart rate) (Budner et al. 2017). As an extension, additional statements were developed to gather data on basic emotions to get a deeper understanding in if and how emotions may affect the HAI. The six basic emotions happiness, stress, uncertainty, superiority, dejection, and excitement are based on Arnold (1960), Darwin (1872), Dornes (1995), Ekman (1999), Johnson-Laird and Oatley (1987), and Plutchik (1962). Superiority is understood as dominant behavior and respect. The basic emotions were inquired by asking the owner to state how much he/she agrees with the statement (0 ¼ I do not agree, 1 ¼ neutral, 2 ¼ I do agree). They were entered manually by the owner via the Pebble Watch before and after every HAI training. The sample of this explorative study included two dogs and two horses (Animal ¼ A) and three owners (O). Data were collected between 28 October 2017 and 30 December 2017 via the Pebble Watch as well as by direct observations and videos in order to evaluate the behavior and success of the different kinds of training sessions described in Sect. 2. Gathered data was analyzed by doing a descriptive statistical analysis as well as a qualitative evaluation based on the theoretical settings. Therefore, quantitative data was exported and transferred into MS Excel and SPS, and qualitative results were mainly analyzed by first documenting all training sessions and observations in an MS Excel sheet and second by drawing an error matrix, in which criteria of measurement quality were documented.

4 Results and Discussion In this section, there will be given an overview of factors that are influencing the measurement quality first. Afterward, explorative results about the owner, the animal, and the connection of both parties are shown. Finally, the challenges and opportunities of this approach are discussed.

170

4.1

K. Stolz et al.

Criteria Influencing Measurement Quality

The following will give an insight into what worked well or not along the experiment. First, dealing with the handling of animals, the best attachment area regarding dogs would have been the thighbone; however, the dogs in this study did not accept this way of attachment. As most dogs living together with humans are familiar with collars, this seemed to be the most convenient alternative. Nevertheless, the data collection quality was bad as the output only showed punctual VMC and no BPM data. This might be due to the thick fur (both dogs had long hair) and the wrinkles. With horses, this problem did not occur, and the most data were received. According to Pebble Support (2016), even dark tattoos may reduce the accuracy and ability of the optical monitor to read the beat of the heart. The experiment showed that the quality of measurement was even higher when the skin of the horse was wet as is the case on a rainy day. Hence, the following findings of this exploratory study on HAI relate to horses. With regards to technical issues, one of the biggest problems was to find suitable smartphones that were able to find the Pebble Watch. Further, there were several problems related to mobile Internet access. Due to the setting of the experiment that most of the time took place outside, the WLAN connection was no option in most cases. Surprisingly, a better data collection was achieved during outside riding sessions compared to working in the round pen or at the same location. In addition, out of 44 times, there was no collection of mood data in 38 times and no sensor data in 26 times. In some cases, it is unclear if there was no measurement of data or a transmission problem. The influencing factors on the measurement quality related to the parties involved in the HAI can be summarized as follows (see Fig. 2), suggesting improvements for better data collection quality in the future:

4.2

Implications from the Data

In this section, there will be a focus on the owner’s data outputs first, before looking at those of the animals (horses) and the interaction of both at the end. Data is based on the exported mood and sensor data from the Pebble Watch in Excel. It is adjusted for incorrect data (due to inappropriate time reporting in comparison to the documented time slots of the trainings), frozen data (as an identical value over a longer time, which did not coincide with the observation by the video documentation), as well as unrealistic data (e.g., heart rate of “190,000” which suggests errors in the data collection). Owner Figure 3 shows the average of the basic emotions over all training sessions before and after the sessions. Analyzing if the owner reports a lower stress level at the end of an HAI compared to the beginning, the experiment shows that the stress level of the owner goes down in the overall analysis. Separating into the different

Measuring Human-Animal Interaction with Smartwatches: An Initial Experiment ANIMAL

HAI

Horse/dog

Settings

 Fur condition (wet, summer fur)  Breed (size, weight)  Mainly manual data input (unreliable algorithm)  Attachment (main artery)

171

HUMAN Owner

 Radio frequency (WLAN, LTE, 4G)  Bluetooth connection (offline data)  Location  Type of smartphone  Weather (e.g. rain)

 Way of attachment (wrist, tightly fastened)  Skin conditions (no tattoos)  Training of the learning algorithm

Fig. 2 Influencing factors on the measurement quality

2.0 1.7 1.5

1.0

0.9

1.0

0.9 0.7

0.6 0.5

0.0

0.3

happiness

stress

1.0

0.7 0.5

0.3

uncertainty

0.2

superiority

dejection

excitement

Fig. 3 Basic emotions of the owner—average over all training sessions (n ¼ 2)

types of training, the stress level remains steady during the round pen session, as the tension whether the animal will accept the owner as the leader continues till the end. Furthermore, as the time of the day for the training sessions, especially for groundwork, usually was after work, the stress level of the owner might have been higher due to work-related stress. The increasing happiness and the decreasing uncertainty and dejection might be other influences on this behavior. As the owner has to enforce his or her will, the dominance level remains stable for all types of training. This is an indicator of a successful HAI and a further reason for the declining stress. Regarding the self-reported pleasance of the owner, especially groundwork resulted in the highest self-reported pleasance at the end of the HAI (see Fig. 4).

172

K. Stolz et al.

Average pleasance (owner) 3

AVERAGE OVERALL TRAINING SESSIONS

2 ROUND-PEN 1

RIDING GROUND WORK

0 AT THE BEGINNING

DURING

AT THE END

Fig. 4 Average pleasance of the owner (n ¼ 2)

One possible reason might be the relaxed atmosphere during groundwork, which can be seen in the low stress level. The pleasance by riding and round pen is also growing to the end of an HAI; however, the basic level of pleasance is at a higher level from the beginning by riding. One cause could be the anticipation of going out with the horse for a longer time. Another reason might be the awareness of the fact that the round pen is more challenging for the owner (see stress and activation). In summary, the statistical analysis determined a highly positive and significant correlation between the variables time and pleasance of the owner with r ¼ .93 ( p ¼ .00)—the longer the exercise went on, the higher the pleasance of the owner—which can be explained by the fact that the variation of time and pleasance is identical overall training sessions. Also, this result has to be considered in relation to the small sample (distortion effects). It was also observed that activation is experienced in the same way overall training session (see Fig. 5). The lowest level of activation during riding is surprising, as the owner’s VMC is at the highest level. One reason could be that while riding, the hands are moving a lot, whereas the body is not that moving. A statistical check in SPSS supports this assumption of a positive relationship between the time (measured by the length of the HAI in minutes) and the heart rate of the owner (BPM O) with r ¼ .42 ( p ¼ .01). Animal Due to the problems in the collection of data with horses (see Sect. 4.1), there are only a few insights on the horses’ behavior based on the observations by the owner, which means that there is a self-reporting bias. However, in specific situations, the following could be observed: first, the rapid increase of the documented heart rate on the Pebble Watch while riding and suddenly being close to a hunting session, linked to an unrestful behavior that might indicate a higher stress level; second, moving heart rate values on the Pebble Watch while feeding treats as a reward; third, the reduction of dominant behavior during round pen and riding sessions; and fourth, the growing signs of pleasance over time in all training

Measuring Human-Animal Interaction with Smartwatches: An Initial Experiment

173

Average activation (owner) 3 AVERAGE OVERALL TRAINING SESSIONS

2 ROUND-PEN 1

0 AT THE BEGINNING DURING

RIDING GROUND WORK AT THE END

Fig. 5 Average activation of the owner (n ¼ 2)

sessions, shown by moving the head down and snorting. Due to the nonsignificant data basis (small N), the animal part was excluded from statistical tests. HAI To analyze the BPM and VMC values, the three types of training with horses had to be distinguished (see Sect. 4.1). Hereby, the values of the basic emotions change in the same way, irrespective of the type of training. It is noticeable that there might be a correlation between the increasing average movement of the animal (measured as VMC) and the decrease of more negatively charged basic emotions of the owner (stress, uncertainty, and dejection) as well as the variation of the average movement of the animal (measured as VMC) and the increase of the positively charged feelings (happiness and excitement) of the owner. Regarding observed data on animals, while riding, the heart rate and the average movement of both the owner and the horses are increasing as this is a typical sequence of a long warm-up, the main high load, and a short cooldown phase. Hence, the higher the movement, the higher the heart rate of both parties during riding. Consequently, the variation of the heart rate can be explained through the movement and the aim during riding (see Figs. 6 and 7). While doing groundwork, the heart rate of the animal is not moving in a similar direction to the movement, which can be explained due to the fact that the animal is expected to stay calm during grooming at the beginning (low movement), whereas the heart rate increases as the horse might be excited due to the expected treats. During the session, the movement intensity grows as there are little exercises, wherefore VMC A is moving up (see Figs. 8 and 9). The heart rate (linear BPM values) of the animal is varying similarly to the owner’s during riding (lines are moving up) and groundwork sessions (lines are moving down), whereas the results are varying inversely during round pen (see Fig. 10). Successful round pen training may be indicated by decreasing need for direct commands through an explicit body language of the owner. The heart rate and the movement of the owner decline while those of the animal increase (see Fig. 11).

174

K. Stolz et al. BPM 160

Riding heart rate over time

140 120

BPM Animal

100

BPM Owner Linear (BPM Animal)

80

Linear (BPM Owner)

60 40

0

8

22

32

40

48

60

MIN

Fig. 6 Heart rate (BPM) over time (riding), n ¼ 4 VMC 8000

Riding movement over time

7000 6000 5000

VMC Owner

4000

VMC Animal

3000

Linear (VMC Owner)

2000

Linear (VMC Animal)

1000 0

0

8

22

32

40

48

60

MIN

Fig. 7 Movement (VMC) over time (riding), n ¼ 4

The average pleasance and activation of the owner are growing (see Fig. 12). The BPM and VMC values of the owner are also slightly decreasing. As the decreasing movement of the owner is not the reason for the increasing activation, the growing VMC of the horse might cause the high BPM O and VMC O values (see Fig. 13). While doing groundwork, the pleasance and activation curves of the owner move in different directions, which might be a consequence of the calm situation at the end

Measuring Human-Animal Interaction with Smartwatches: An Initial Experiment

175

BPM 180

Groundwork heart rate over time

160 140

BPM Animal

120

BPM Owner

100

Linear (BPM Animal)

80

Linear (BPM Owner)

60 40

0

10

20

30

MIN

Fig. 8 Heart rate (BPM) over time (groundwork), n ¼ 4 VMC 8000

Groundwork movement over time

7000 6000 5000

VMC Animal

4000

VMC Owner

3000

Linear (VMC Animal)

2000

Linear (VMC Owner)

1000 0

0

10

20

30

MIN

Fig. 9 Movement (VMC) over time (groundwork), n ¼ 4

of the session (e.g., stretching, massages), whereby the pleasance rises enormously (see Figs. 14 and 15). One reason might be that groundwork usually takes place during the week after work when the owner is stressed and the relaxing atmosphere contributes to their pleasance. While riding sessions (see Figs. 16 and 17), the activation and pleasance curves do not increase that much compared to the other types of training sessions. Further,

176

K. Stolz et al.

BPM 180

Round pen heart rate over time

160 140

BPM Animal

120

BPM Owner

100 80

Linear (BPM Animal)

60

Linear (BPM Owner)

40

0

5

8

12

MIN

Fig. 10 Heart rate (BPM) over time (round pen), n ¼ 4 VMC 12000

Round pen movement over time

10000 8000

VMC Animal

6000

VMC Owner

4000

Linear (VMC Animal) Linear (VMC Owner)

2000 0

0

5

8

12

MIN

Fig. 11 Movement (VMC) over time (round pen), n ¼ 4

the pleasance level is higher at the beginning. This might be due to the excitement of going out. As the activation curves move equally to the pleasance curves during round pen sessions (see Figs. 12 and 13) and while riding (see Figs. 16 and 17), there could be a connection between activation and pleasance.

Measuring Human-Animal Interaction with Smartwatches: An Initial Experiment

177

ROUND PEN (n=4) BPM A Animal

180

2

140

1.5

100

1

60

0 BPM A

5 Pleasance O

10

Pleasance/ Activation O Owner

0.5 MIN

- - - Activation O

Fig. 12 Heart rate (BPM) of the animal (A) in context with activation and pleasance of the owner (O)—round pen ROUND PEN (n=4) VMC A Animal

2

5000 4000

Pleasance/ Activation O Owner

1.5

3000 2000

1

1000 0

0 VMC A

5 Pleasance O

0.5 MIN 10 - - - Activation O

Fig. 13 Movement (VMC) of the animal (A) in context with activation and pleasance of the owner (O)—round pen

4.3

Challenges and Opportunities

Due to the pre-experimental design of this paper, there are some limitations and inspirations for future research. As Pebble (2016) admits, the described experiment shows that the measured data was not always accurate. In several cases, the data was “frozen” (see Sect. 4.2),

178

K. Stolz et al. GROUNDWORK (n=4) BPM A Animal

180

2

140

1.5

100

1

60

0

10

BPM A

Pleasance O

20

0.5

Pleasance/ Activation O Owner

MIN

- - - Activation O

Fig. 14 Heart rate (BPM) of the animal (A) in context with activation and pleasance of the owner (O)—groundwork

VMC A Animal

GROUNDWORK (n=4)

5000

2

4000

Pleasance/ Activation O Owner

1.5

3000 2000

1

1000 0

0

VMC A

10 Pleasance O

20

0.5 MIN

- - - Activation O

Fig. 15 Movement (VMC) of the animal (A) in context with activation and pleasance of the owner (O)—groundwork

which means that there were the same values over a longer period of time—even though different paces were documented in the qualitative analysis for the same time span. Second, due to the technical issues mentioned in Sect. 2, the focus was on two horses and their two owners. As the basic emotions are only collected twice, at the beginning and the end of a training session, the findings might only indicate trends.

Measuring Human-Animal Interaction with Smartwatches: An Initial Experiment

179

RIDING (n=4) BPM A Animal

180

2

140

1.5

100

1

60

0

20

BPM A

40 Pleasance O

60

0.5

Pleasance/ Activation O Owner

MIN

- - - Activation O

Fig. 16 Heart rate (BPM) of the animal (A) in context with activation and pleasance of the owner (O)—riding RIDING (n=4) VMC A Animal

2

5000 4000

Pleasance/ Activation O Owner

1.5

3000 2000

1

1000 0

0 VMC A

20

40 Pleasance O

60

0.5 MIN

- - - Activation O

Fig. 17 Heart rate (BPM) of the animal (A) in context with activation and pleasance of the owner (O)—riding

Even if the limited (quantitative) data allows no generalization and is not representative, the results show tendencies that might be the basis for further research. Besides looking for a more objective way of recording the main feelings, future studies might document the basic emotions more frequently and include other factors (moderating variables), like the character of the participants, time of day, or other activities, done before and after the session. Another extension of this study could be

180

K. Stolz et al.

Table 1 Challenges and opportunities Opportunities • Attachment and usage of a Pebble Watch on horses • Supporting positive effects of HAI regarding related work • Separation of the training types of horses (related to aims)

Challenges • Way of attachment and tracking on animals • Reliable data reporting and transmission via the Pebble Watch and other technique • Reliability of the learning algorithm (esp. on horses)

Further research • Deeper insights in body signals while/after the HAI regarding animals and owners • Larger and more diverse sample (esp. A&O pairs, other breeds) • A longer period of time to investigate the HAI

to use machine learning for higher prediction accuracy and using other techniques for measuring animal body signals. Table 1 lists challenges and opportunities.

5 Outlook It can be concluded that the first aim of the project was partially met. The Pebble Watch could not be used for dogs; however, it worked well on horses. The data indicates that there might be connections between the measured Pebble Watch values of the heart rate (BPM), average movement (VMC), and the mood data (basic emotions, pleasance, and activation) of an animal and its owner during training sessions. However, only the owner’s values could be tested statistically with the result that there is a correlation between the progressive time of the HAI and the growing pleasance of the owner as well as the decreasing heart rate of the owner. Further, the variation of the animal’s heart rate changes in synch with those of the owner in groundwork and riding sessions, whereby it is inversely related during a round pen session, which can be explained through the different aims of the three types of training. According to Gehrke (2010) and Beetz et al. (2012), the experiment also supports the relationship between the animals and owners’ HRV frequency, though the type of training session has to be considered. Further, the self-reported stress by the owner is lower at the end of the HAI, except after a round pen session (consistent stress level). One reason for this might be that tension remains high to the end whether the horse accepts the owner as the leader. Hence, this paper confirms the findings of Allen, K. et al. (2002). Furthermore, the relationship between the variable time and pleasance of the owner could be confirmed, especially during groundwork (see Góngora and Solano 2015; Wells 2009). A cause might be that groundwork takes place in a calm environment, in which the animals’ and owners’ heart rate (BPM) and movement (VMC) values go down. To overcome especially the issue of getting a huge volume of data (larger sample and a longer period of time), further studies should use a watch that is designed to measure the body signals of an animal. Therefore, the POLAR Equine might be an

Measuring Human-Animal Interaction with Smartwatches: An Initial Experiment

181

option for the measurement on horses, because it is equipped with synthetic electrodes, which adapt to the movements of a horse and achieve high measurement accuracy due to the permanent contact to the skin of the horse by attaching the sensor under the saddle pad (Polar Electro 2017a). As the horse might be more relaxed when knowing the rider well (trust), it would be interesting to measure data with animal-owner pairs that are not used to each other (e.g., during the break-in of a horse). In addition, other circumstances (e.g., summer fur, shorthair dogs) might lead to a data collection on dogs. Summarizing the initial experiment on HAI, the following are the main lessons learned and proposals for future research: 1. The Pebble Watch is able to measure body signals on animals. However, a further study using specialized animal tracking tools would improve the results. 2. Regarding the insights into the HAI by using the measured data, the practical relevance would be greater, if the H-A pairs would not know each other that well or if there are cases of singularity, e.g., illness. Therefore, future research should increase the sample size and sample diversity. 3. Personal traits might influence the HAI. Thus, a more detailed analysis of the basic emotions in combination with the Big Five personality test of the animal (without having the owner as the appraising person) would gain deeper insights. 4. Specializing the learning algorithm on horses (or animals in general) might be another progress to ensure that measurable surroundings (e.g., windiness, temperature) relate to the happiness of the horse (not the owner).

References Allen, K. M., Blascovich, J., Tomaka, J., & Kelsey, R. M. (1991). Presence of human friends and pet dogs as moderators of autonomic responses to stress in women. Journal of Personality and Social Psychology, 61(4), 582–589. Allen, K., Blascovich, J., & Mendes, W. B. (2002). Cardiovascular reactivity and the presence of pets, friends, and spouses: The truth about cats and dogs. Psychosomatic Medicine, 64, 727–739. https://doi.org/10.1097/01.PSY.0000024236.11538.41. Arnold, M. B. (1960). Emotion and personality. New York: Columbia University Press. Barker, S. B., Knisely, J. S., McCain, N. L., & Best, A. M. (2005). Measuring stress and immune response in healthcare professionals following interaction with a therapy dog: A pilot study. Psychological Reports, 96, 713–729. Beetz, A., Uvnäs-Mobergm, K., Julius, H., & Kotrschal, K. (2012). Psychosocial and psychophysiological effects of human-animal interactions: The possible role of oxytocin. Frontiers in Psychology, 3, 1–15. https://doi.org/10.3389/fpsyg.2012.00234. Brown, J. S. (2001), Brown, Joseph S. US 6,311,645 B1. Budner, P., Eirich, J., & Gloor, P. A. (2017) “Making you happy makes me happy”: Measuring individual mood with smartwatches. Retrieved from https://arxiv.org/ftp/arxiv/papers/1711/ 1711.06134.pdf Budzinski, O., & Schneider, S. (2017) Smart fitness: Ökonomische Effekte einer Digitalisierung der Selbstvermessung. Darwin, C. (1872). The expression of the emotions in man and animals. London: John Murray.

182

K. Stolz et al.

Deine Tierwelt GmbH & Co. KG. (2015) Hundeerziehung – Grundausbildung: Auf dem Weg zum guten Hund. Retrieved from https://www.deine-tierwelt.de/magazin/hundeerziehunggrundausbildung-auf-dem-weg-zum-guten-hund/ Dornes, M. (1995). Gedanken zur frühen Entwicklung und ihre Bedeutung für die Neurosenpsychologie. Forum Der Psychoanalyse, 11, 27–49. Ekman, P. (1999). Basic emotions. In T. Dalgleish & M. Power (Eds.), Handbook of cognition and emotion (pp. 45–60). Sussex: Wiley. Evans, D. L. (2000). Training and fitness in athletic horses: A report for the rural industries research and development corporation. University of Sydney. Gehrke, E. K. (2010). The horse-human heart connection: Results of studies using heart rate variability. NARHA’s STRIDES, 20–23. Góngora, V. C., & Solano, A. C. (2015). Psychometric properties of the three pathways to well-being scale in a large sample of Argentinean adolescents. Psychological Reports, 117(1), 167–179. Johnson-Laird, P. N., & Oatley, K. (1987). Towards a cognitive theory of emotions. Cognition and Emotions, 1, 29–50. Pebble. (2016). Pebble 2. Retrieved from https://www.pebble.com/pebble-2-smartwatch-features Pebble Support. (2016). Heart rate monitor introduction. Retrieved from https://help.getpebble. com/customer/en/portal/articles/2561918-heart-rate-monitor-introduction?b_id¼14173 Pebble. Pebble health. Retrieved from https://developer.pebble.com/guides/events-and-services/ health/ Pferdeflüsterei. (2017a). Fein! Reiten. Retrieved from https://www.pferdefluesterei.de/fein-reiten/ Pferdeflüsterei. (2017b). Warum Bodenarbeit wichtig ist für die Pferd-Mensch-Beziehung. Retrieved from https://www.pferdefluesterei.de/a-z/bodenarbeit/ Pferdeweb. (2017). Bodenarbeit im Roundpen. Retrieved from http://www.pferdeweb.ch/ bodenarbeit-im-roundpen-12.html Plutchik, R. (1962). The emotions: Facts, theories and a new model. New York: Random House. Polar Electro. (2017a). Elektroden-Basisset für Equine H3 Herzfrequenz-Sensor. Retrieved from https://www.polar.com/de/produkte/equine/accessoires/elektroden_basisset_f%C3%BCr_ equine_h3_herzfrequenz-sensor Polar Electro. (2017b). Gesundheit, Fitness und Leistung des Pferdes. Retrieved from https://www. polar.com/de/produkte/equine/gesundheit_fitness_und_leistung_des_pferdes Posner, J., Russell, J. A., & Peterson, B. S. (2005). The circumplex model of affect: An integrative approach to affective neuroscience, cognitive development, and psychopathology. Development and Psychopathology, 17(3), 715–734. Quantified Self Labs. (2011). Guide to self-tracking tools. Retrieved from http://quantifiedself.com/ guide/ Statista. (2018). Fitness tracker device sales revenue worldwide from 2016 to 2022 (in billion U.S. dollars). Retrieved from https://www.statista.com/statistics/610433/wearable-healthcaredevice-revenue-worldwide/ Wells, D. L. (2009). The effects of animals on human health and well-being. Journal of Social Issues, 65(3), 523–543.

Show Me Your Moves: Analyzing Body Signals to Predict Creativity of Knowledge Workers Marius Stein, Peter A. Gloor, and Daniel Oster

Abstract We propose a novel approach to measuring the collaboration of knowledge workers, using body sensing smartwatches to capture psychometric data about individuals in a team. In a proof of concept study, we collected 2653 samples of body signals by equipping 15 people with our body sensing smartwatch over the course of 3 days during a design workshop. Additionally, we polled the users about their selfperceived team creativity at the end of each day. By employing multiple linear regression models, we found that body signals tracked by the smartwatch correlate significantly with the perceived team creativity reported by the individuals. Comparing those correlations with known predictors of creativity such as mood states and personality traits, we found that movement-related body signals predict creativity on the same accuracy level as mood states and personality traits do.

1 Introduction The ability to innovate is one of the most important competitive advantages for companies (Somech 2006). In order to survive, companies constantly compete to deliver the most creative and innovative new products, to adopt to changing governmental regulations or to adjust to rapidly shifting market conditions – all of which require creativity (Amabile 1988). In order to succeed in this never-ending battle of adaption, companies rely on creative individuals working together in productive work groups. Social science and psychology provide the foundation for studying human collaboration. Nonetheless, the output of collaborating knowledge workers is still

M. Stein (*) · D. Oster University of Cologne, Cologne, Germany e-mail: [email protected] P. A. Gloor MIT Center for Collective Intelligence, Cambridge, MA, USA e-mail: [email protected] © Springer Nature Switzerland AG 2019 Y. Song et al. (eds.), Collaborative Innovation Networks, Studies on Entrepreneurship, Structural Change and Industrial Dynamics, https://doi.org/10.1007/978-3-030-17238-1_11

183

184

M. Stein et al.

hard to measure. For decades scientists and practitioners relied mostly on sparse, qualitative, or subjective observations like survey responses or observed behavior in order to measure workgroup performance and creativity (Aral et al. 2012; Daggett et al. 2017). However, physiological data about users can likely provide “a richer account of user cognition than that obtained from any other source, including the user himself” (Minnery and Fine 2009, p. 73). This assumption is reflected in the vast amount of literature that examines cognition and especially creativity by means of neuroscientist methods (e.g., Gaskin et al. 2017; Jausovec and Bakracevic 1995; Léger et al. 2014; Wagner et al. 2005). Most studies utilizing methods of neuroscience to study creativity and cognition use medical grade equipment to collect body signals from participants (e.g., Léger et al. 2014; Wagner et al. 2005). While this equipment is capable of collecting body signals with high precision, the costs of such equipment are also quite high. Gaskin et al. (2017), for example, claim that the cost for professional heart rate monitors, EEG devices, and blood oxygen sensors is multiple thousands of dollars. Another issue with high-cost equipment is that it constraints the sample size as can be seen by various studies with low sample size in this area (e.g., Jausovec and Bakracevic 1995; Léger et al. 2014). Besides high cost and low sample sizes, medical grade equipment usually requires the user to be heavily tethered in order to measure body signals. The tethering of participants, however, might confound the actual measures of cognition and creativity. Recent advances in digital technology, namely, the rise of body sensing smartwatches, enable us to overcome all three of the limitations outlined in the paragraph above. Modern smartwatches can collect different physiological measures about its wearer (e.g., heart rate, blood oxygen levels, acceleration, lighting) at a fraction of the cost for medical grade equipment. Additionally, a smartwatch is as unobtrusive as a bracelet (Gaskin et al. 2017), which minimizes the effect of the measurement device on the users’ behavior. In this exploratory proof of concept study, we aim to evaluate the use of body sensing smartwatches to predict creativity of knowledge workers in organizational settings. The research question that we are answering is: Can the data collected by smartwatch-based body sensors be utilized to identify factors that foster or mitigate creativity?

2 Related Work 2.1

Honest Signals

The theoretical idea of this work is that human action and the reasoning for such action is mirrored in physiological and social signals. When we interact with individuals, those signals can display an objective measure of how fruitful the interaction is. Pentland (2010) called those signals honest signals in their 2010 book where they elaborated on implications, examples, and opportunities that arise from the availability of such signals.

Show Me Your Moves: Analyzing Body Signals to Predict Creativity of. . .

185

How those honest signals can be used to quantify human interaction is best demonstrated by an example: during one study Pentland (2010) invited a group of business executives to present one business plan each. Afterward, each presentation was peer evaluated, and the business plans were ranked according to how well the group thought the idea would sell. Additionally, each presenter was equipped with a sociometric badge to measure different social signals. The badges captured the variability in speech energy, the movement of the presenters, and how many back and forth gestures such as smileys and head nodes occurred between the presenter and the audience. When using these measures to predict which business plan the executives would rank highest, Pentland (2010) achieved an almost perfect accuracy. In another study Pentland (2010) were able to use honest signals to predict job satisfaction and even creativity. Pentland (2010) explicitly elaborate on four different honest signals that the sociometric badge captured in the example above: Influence Is the amount of influence one person has over another during a social interaction. It can be measured by how much each interaction participant was talking. If one person was talking more than her/his counterpart, she/he had a higher influence during the interaction. Mimicry Pentland (2010) argue that when two people are deeply engaged in a conversation and on the same wavelength, then they tend to copy each other’s actions. We can measure the amount of mimicry in a social interaction by either analyzing the speech patterns of both participants or by taking a look at the physical activity. The more alike the speech patterns and the physical activity are, the higher the degree of mimicry. Activity The intensity of physical movement or speech energy during a social interaction. If we are excited about a conversation topic or simply interested in our conversation partner, we tend to talk more energetically and move more. Consistency Consistency is an indicator of how consistent our speech patterns or movements are during a social interaction. If our speech pattern or our movement intensity changes during an interaction, then consistency is low. Low consistency can occur when a person is nervous, and high consistency might be an indicator of high self-confidence. Within our research we measure activity and consistency of participants wearing the body sensing smartwatch application.

2.2

Creativity

In this paper we understand creativity as the production of novel and useful ideas by an individual or small group of individuals working together (Amabile 1988). During an interview study, Amabile (1988) identified factors that promote individual

186

M. Stein et al.

creativity based on the qualities of the problem solvers in the group. She found that among others especially mutual understanding, the quality of the group (whether people can work together efficiently) and the social skills of the group members are paramount properties that influence the creative work of the group. For the creative process, Glover et al. (2013), following Yamamoto’s (1963) work, outlined different properties of the creative thinking process. They argue that creativity is expressed in the dimensions flexibility (many different ideas), fluency (lots of ideas), inventiveness (inventing and developing ideas), originality (unique ideas), and elaboration (detailed ideas).

2.3

Mood and Creativity

Despite the qualities of the problem solver as factors that influence creativity, the individual mood stands out as the most widely studied and least disputed predictor of individual creativity (Baas et al. 2008; Isen and Baron 1991; Mumford 2003). The reason for the popularity of the mood state as predictor of creativity is due to the fact that mood state is seen as intermediary between personality and situational-related predictors, on the one hand, and creative performance, on the other hand (Baas et al. 2008). In other words, personality and situational effects are mirrored in a person’s mood which enabled us to use mood as an intermediary predictor to creative performance. In this paper we understand the mood-creativity link as a two-dimensional relationship. The first dimension is defined by the relationship between positive mood (in contrast to a neutral mood state) and creative performance. This relationship is based on a rich body of knowledge as many researchers have found that a positive mood state results in better creative performance of an individual (Ashby and Isen 1999; Lyubomirsky et al. 2005). However, there are some exceptions to these findings as researchers have also found situations in which individuals in a positive mood are less creative than persons in neutral mood states (Vosburg 1997). The second dimension of the mood-creativity relationship is defined by the relationship between negative mood and creative performance. The influence of negative mood on creative performance is highly disputed in the scientific literature. While some researchers have found a positive link between negative mood state and creativity (e.g., Adaman and Blaney 1995; Clapham 2001), others have found the opposite (e.g., Mikulincer et al. 1990; Vosburg 1998) or even no relationship at all (e.g., Göritz and Moser 2003; Verhaeghen et al. 2005).

2.4

Personality Traits

Personality traits are understood as characteristics of an individual’s behavior. Personality traits are assumed to be stable over time yet can be changed in systematic

Show Me Your Moves: Analyzing Body Signals to Predict Creativity of. . .

187

ways (Conley 1985; Robins et al. 2001). In psychological research Norman (1963) laid the foundation for today’s research in personality by introducing a taxonomy that is based on five distinct personality traits. These five distinct personality traits, today known as Norman’s five (or big five), have shown a robustness in recent years as they have been validated multiple times in different contexts (Bond et al. 1975; Conley 1985; Digman 1990; Digman and Inouye 1986). The five traits are: 1. Extroversion: indicates how outgoing an individual is. Persons that have a high score for extroversion are talkative, adventurous, and usually seek stimulation in the company of others. 2. Agreeableness: displays the desire for social harmony. Agreeable people usually get along well with other people and are more collaborative and good natured. 3. Consciousness: individuals that score high on the consciousness personality trait are more tidy and dutiful. 4. Emotional stability (anxiety/neuroticism): is a personality trait that defines how anxious an individual is. Persons with a high emotional stability usually can handle stress better and are calmer and less nervous. 5. Culture (openness to new experiences): is a personality trait that influences the curiosity of people. Open people are more interested in getting to know new cultures, have a high appreciation for art, and are more likely to accept unusual ideas. There is a vast body of research about the influence of these personality traits on job performance (Barrick and Mount 1991) and creativity (Sung and Choi 2009).

3 Research Methodology 3.1

Hypothesis

In order to answer the question whether physiological measures collected by a smartwatch reflect self-perceived creativity, we developed the following hypothesis: Body signals captured by a smartwatch correlate significantly with an individual’s perceived creative performance measured by self-evaluation.

We will run a statistical analysis with data collected from a study as explained in the remainder of this section. Since personality traits and mood state are well-known to influence the creativity of an individual (Baas et al. 2008; Isen and Baron 1991; Mumford 2003; Sung and Choi 2009), we use those as a baseline to compare the strength of the correlation between body signals and self-perceived creativity against. We assume the correlation between body signals and self-perceived creativity to be equally strong as the correlation between mood states and self-perceived creativity or personality traits and self-perceived creativity. The overall goal of this

188

M. Stein et al.

research is to explore the possibility of utilizing smartwatch-based body sensors in creativity research as a substitute for medical grade equipment.

3.2

Measuring Creativity, Mood States, and Personality

We are developing a smartwatch application called Happimeter1 in order to measure a person’s honest signals and happiness over time and compare this with selfreported creativity. The Happimeter is an application that was developed at the MIT Center for Collective Intelligence. In order to measure the mood state, the Happimeter uses an experience-based sampling method. The user is polled four to eight times a day at random times to rate her personal happiness (Budner et al. 2017). Following Amabile’s and Glover’s characterization, we designed the following five question survey for participants to self-evaluate their creativity. For each of the items, we gave a Likert-type scale as answer option with which the participants could indicate if they either (a) totally disagree, (b) somewhat disagree, (c) neither disagree nor agree, (d) somewhat agree, or (d) totally agree with the given item. The five questions of the inventory are the following: (Q1) Would you characterize today’s teamwork as successful? (Q2) Would you characterize today’s group work as creative in the sense that you had many different ideas? (Q3) Would you characterize today’s group work as productive in the sense that the ideas of your group where of high quality? (Q4) Would you agree that your team members motivated you? (Q5) Was there fair and equal participation of every team member in your group? (Q1) aims to assess the mutual understanding within the specific group. If a participant highly agrees with (Q1), it indicates that the mutual understanding within the group is good. (Q2) and (Q3) assess the perceived creativity of the individuals within the group based on Glover et al. (2013) and Yamamoto’s (1963) characterization. (Q2) is also tailored to measure the flexibility and the fluency of the creative thinking process. (Q3) assesses inventiveness, originality, and elaboration of the creative thinking process. (Q4) measures the quality of the group in which the creative process takes places, and (Q5) assesses the social skills of the group members. The participants of the study were instructed about the meaning of the different questions to guarantee a shared understanding regarding the survey. To get a quantified measure of creativity, each of the questions was given a score depending on the answer given by the participants. If a participant answered with

1 The Happimeter smartphone and watch code is open source and free for academic use. More information can be found at https://happimeter.org

Show Me Your Moves: Analyzing Body Signals to Predict Creativity of. . .

189

“highly disagree,” then the score was 1, if the participant answered with “highly agree,” then the score was 5. The scores of each question were added up to get a single measure of creativity per participant. This single measure of creativity has a value between 5 (least creative) and 25 (highly creative). Although using self-rating of creativity is disputed by some researchers, others have obtained robust findings utilizing such a method (Sung and Choi 2009). The problem with this approach is the subjectivity of such an assessment. Individuals might have different understandings of creativity, and some might rate the same outcome more positively, while others do it more negatively. However, we designed our questionnaire to minimize those effects by clearly elaborating and differentiating the different items of the questionnaire. Additionally, we gave basic instructions to the participants on how to understand and rate each of the items within the inventory. Therefore, we are confident that we could reduce the subjectivity issue of self-rating to a minimum. To sample the mood state of a user, a two-dimensional model, similar to the circumflex model of affect, is utilized with one dimension being pleasance and the other dimension being arousal (Posner et al. 2005; Russell and Pratt 1980). For each of the two dimensions, the user can indicate on a three-point scale how strongly aroused or how strongly pleasant (s)he feels. Despite the popularity of using extensive, multi-item inventories to measure happiness (e.g., Hills and Argyle 2002), studies have shown that short scales can be as valid as long ones when measuring happiness of individuals and lengthening inventories beyond a certain point might even decrease the validity (Bell and Lumsden 1980). Additionally, measuring happiness using a single-item scale has been demonstrated to be reliable and valid (Abdel-Khalek 2006). The benefit of measuring happiness in the way we do it is that we can ask the same questions multiple times a day without taking away too much time from the user. Usually, answering the two questions on the smartwatch does not require the user’s attention for more than a few seconds. This ensures that users are not interrupted in their daily routine and the effects of the measurement on the user’s behavior are reduced to a minimum. To measure an individual’s personality, we used the NEO Five-Factor Inventory (NEO-FFI) (Costa and McCrae 1989). The NEO-FFI comes in different versions, which differ in the number of questions they contain. The version utilized during our experiment had 50 questions. For each of the five personality types (neuroticism, openness, extroversion, consciousness, and agreeability), there are ten questions that assess the extent to which one person inherits that specific personality type. To each question a four-point Likert scale is given as answer option. As a result of the test, each of the five personality traits has a score between 0 and 30 points. A score of 0 means that the individual does not inherit that given personality trait at all; a score of 30 indicates that the individual inherits the personality trait to a high degree. The consistency of the NEO-FFI ranges from 0.68 to 0.86 which proves its validity (McCrae and Costa 2004). Figure 1 summarizes the relationship between the different constructs relevant to this study.

190

M. Stein et al.

BODY SIGNALS

The relationship between Mood States and Creativity as baseline

MOOD STATE

The relationship between Personality Traits and Creativity as baseline

PERSONALITY TRAITS

Baas et al., (2008)

CREATIVITY

Am

ab

ile

(1 9

88

)

H: Body signals captured by a smartwatch correlate significantly with an individual’s perceived creative performance measured by self-evaluation.

Fig. 1 Body signals, mood states, and personality traits

3.3

Experimental Setting

To collect data, we conducted a study at a German bank. The study was conducted over the course of 3 days in which knowledge workers from several subsidies of the company were invited to participate in a workshop. The goal of the workshop was to spread a creative mindset within the company by educating key employees. The workshop consisted of several short, lecture-like sessions followed by interactive sessions in which the participants were advised to solve problems by employing creative methods. The agenda of the workshop was loosely oriented by the different phases of the design sprint method—a method developed by Google to nurture creativity when solving problems (Banfield et al. 2015). In total 28 employees participated in the workshop. Because of hardware availability issues, only 15 participants could be equipped with our body sensing smartwatch that had the Happimeter app installed to track body signals and mood states. The participants were selected randomly and had the opportunity to skip participation. These 15 participants were split up into 3 groups with each group working on different problems. In order to track the body signals, the Happimeter application measured the different body signals in 15-min intervals and reported them back to our database. To track the mood states, the Happimeter app asked the participant to indicate her mood state four to eight times a day. At the end of each day, all participants were asked to fill out the creativity survey to indicate their perceived creativity. Additionally, each participant was asked to fill out a NEO-FFI personality test. During the introduction phase of the workshop, the participants were briefed about the different surveys in order to ensure a common understanding of the

Show Me Your Moves: Analyzing Body Signals to Predict Creativity of. . .

191

different items within the surveys. Additionally, the participants were introduced to the research goals and the method of data collection. After the introduction phase, the participants were asked whether they wanted to participate in the study.

3.4

Hardware and Software

In our experiments we used a Pebble 22 smartwatch to gather the different body signals. The Pebble 2 integrates the following sensors that are accessible via the SDK: heart rate monitor, three-axis accelerometer, and an ambient light sensor. Utilizing these sensors, we tracked six distinct body signals of the participants during our study. Prevailing Activity Within the Last 15 Minutes (Activity) The activity can either be resting/sleeping, unspecific, walking, or running. Pebble uses the measures of the sensor to identify the current activity of its wearer. A detailed explanation of the algorithms used to discern different activities was published by (Stockham 2016). Average Heart Beats per Minute Within the Last 15 Minutes (Avg. bpm) Studies have shown that stress and health conditions are mirrored in a person’s heart rate (Thayer et al. 2012), which we assume to be influencers of creative performance. Additionally, prior research has shown that the heart rate reflects an individual’s perceived creative performance (Jausovec and Bakracevic 1995; Sternberg 1999). Average Surrounding Light Level Within the Last 15 Minutes (Avg. Light Level) The avg. light level variable is an indicator for the intensity of the surrounding light. It is taken from the ambient light sensor of the Pebble smartwatch and can have a value between one and five. Different lighting conditions are known to influence happiness and creativity (Knez 1995). Average of Accelerometer Intensity Within the Last 15 Minutes (Avg. Accelerometer) This body signal is an indicator for the magnitude of movement in the physical space. It is calculated by summing up the average values of the three axes of the accelerometer. Studies have shown that more rigorous movement and especially sports can have an impact on an individual’s creative performance (Bowers et al. 2014). The acceleration is measured in milli-G (gravitational force). Variance of the Accelerometer Intensity Within the Last 15 Minutes (Var. Accelerometer) This body signal is an indicator for the consistency of movement in the physical space.

2

Unfortunately, the pebble 2 smartwatch has been discontinued and is not produced anymore. The Happimeter was migrated to a generic Android platform. More information can be found at https:// happimeter.org

192

M. Stein et al.

Vector Magnitude Count During the Day (VMC) VMC is a measure of the total movement over the course of the day. More rigorous movement yields higher VMC values (Stockham 2016). We argue that those body signals reflect the same honest signals that Pentland (2010) used to predict the performance of presenters as introduced in the previous section. The honest signal activity is reflected in the heart rate and the movement (VMC, acc. avg., and activity) of individuals. The accelerometer variance is a measure of consistency. In comparison to the aforementioned sociometric badges, the Pebble smartwatch provides the possibility to measure heart rate, light level, and the activity through accelerometer. Additionally, the sociometric badges are able to make audio recordings and measure physical proximity between people. These features are currently being added to the new Android watch version of the Happimeter.

4 Results In order to test our hypotheses, we employed different statistical methods such as correlation analysis between the creativity score and the independent variables and different multilevel regression models.

4.1

Data

We collected 2653 valid observations of body signals through the smartwatch application. Additionally, all 15 participants filled out the creativity survey at least once with a total of 34 surveys that were filled out. Moreover, 298 pairs of pleasance-arousal samples could be collected through the smartwatch application. In order to benefit from the large number of collected body signals, the pleasance-arousal samples and the filled-out surveys were related to each of the 2653 collected body signals in such a way that one pair of body signals has exactly one pair of pleasance-arousal samples and one creativity score assigned to them. The first two variables, pleasance and arousal, represent the mood state of a person. For the pleasance variable, no participant reported an unpleasant mood state during the time of our experiment. Additionally, 225 instances of neutral pleasance and 2428 instances of high pleasance were reported. Regarding the arousal variable, we collected 607 instances of a low arousal state, 1904 instances of a neutral arousal state, and 143 instances of a high arousal state. Values for the average (heart)-beats-per-minute (avg. bpm) variable vary between 0 and 239 bpm. Values below 20 bpm and above 220 bpm are considered measurement errors and were replaced by the overall mean. The average light level (avg. light level) variable ranges from one to five, with increasing values representing increasing levels of surrounding light. Values for VMC, acc. avg., and acc. var. were

Show Me Your Moves: Analyzing Body Signals to Predict Creativity of. . .

193

Table 1 Independent variables Variable High pleasance Neutral pleasance Low pleasance High arousal Neutral arousal Low arousal Avg. bpm Avg. light level Activity: rest Activity: none

# Valid measures 2428 225 0 143 1904 607 2653 2653 1097 1479

Variable Activity: walk Activity: run VMC Acc. avg. Acc. var. Is open Is agree Is conscience Is extro

# Valid measures 114 3a 2653 2653 2653 2 2 2 2

The three measures of “Activity: run” all resulted from one person that went for a 45-minute run in the morning of one workshop day

a

scaled by the root-mean-square of the series. The values for VMC range from 0 to 1.6545, the values for acc. avg. range from 0 to 1.3640, and the values for acc. var. range from 0 to 5.9208. The last four variables represent the different personality types. We coded the different personality types as binary variable; a value of 1 indicates that a participant inherits a personality trait to a high degree (more than 25 points in the NEO-FFI inventory). Within our study we found no participant that inherits a personality trait that is related to anxiety. Table 1 summarizes all independent variables. The dependent variable creativity score ranges from 14 points to a maximum of 25 points with a mean of 20.12 points. This distribution indicates that no participant reported having an extremely low self-perceived creativity. On the contrary, 25 out of 2653 measures indicate a maximum value of 25 for self-perceived creativity.

4.2

Body Signals and Creativity

One major finding was that movement-related body signals do correlate with the selfperceived team creativity. We found that especially the movement intensity, measured by the accelerometer averages, does correlate significantly with self-perceived creativity. Additionally, the activities of the participants, measured by the Pebble activity metric, demonstrate an even better predictor for creative performance. Lastly, also the heart rate average does correlate significantly with creative performance. A regression analysis with body signals as independent variables further strengthens our argument that self-perceived creativity is mirrored by the body signals. We find that the movement intensity measured by the accelerometer average has a positive impact on creativity, whereas a high variability in movement measured by the accelerometer variance has the contrary effect. When looking at the different activities that are tracked by the smartwatch, we find that if a person went for a run before the workshop started, she/he perceived her/his work more creative.

194

M. Stein et al.

These results demonstrate that body signals captured by a smartwatch can indeed reveal the self-perceived creativity of a person. Especially the movement patterns of individuals and whether a person is doing sports before a working session seem to predict the creativity of individuals.

4.3

Mood States, Body Signals, and Creativity

To validate the results shown above, we compared the correlations between body signals and creativity with the correlation between mood states and creativity (Tables 2, 3, and 4). We found that self-perceived creativity correlates with arousal as well as with movement and the heart rate and also with the activity of a person. Regarding pleasance we did not find a significant correlation. For the multilevel regression (Table 5) with participants as the second-level grouping variable, we first used only the mood state-related variables as predictor for the creativity score. In the second and third model, we employed the same Table 2 Correlation of creativity with mood states (n ¼ 2653)

1 Creativity score 2 Pleasance 3 Arousal

1 1.00 0.02 0.08***

2

3

1.00 0.07***

1.00

***99% Table 3 Correlation of creativity with body signals (n ¼ 2653) 1 Creativity score 2 Avg. bpm 3 Avg. light level 4 Activity 5 VMC 6 Acc. avg. 7 Acc. var.

1

2

3

4

5

6

7

1.00 0.05** 0.02 0.20*** 0.01 0.09*** 0.02

1.00 0.20*** 0.37*** 0.19*** 0.20*** 0.52***

1.00 0.20*** 0.09*** 0.01 0.12***

1.00 0.22*** 0.03 0.31***

1.00 0.04** 0.13***

1.00 0.12***

1.00

***99%, **95% Table 4 Correlation of creativity with personality (n ¼ 2653) Creativity score 1 Is open 2 Is agree 3 Is conscience 4 Is extro ***99%, **95%

1 1.00 0.05** 0.02 0.20*** 0.01

2

3

4

5

1.00 0.20*** 0.37*** 0.19***

1.00 0.20*** 0.09***

1.00 0.22***

1.00

Show Me Your Moves: Analyzing Body Signals to Predict Creativity of. . .

195

Table 5 Multilevel regression models (grouping by participant ID) Variables High pleasance Neutral arousal High arousal Avg. bpm Avg. light level Activity: none Activity: walk Activity: run VMC Acc. avg. Acc. var. Is open Is extro Is conscience Is agreeable Constant Random effects std. dev. R2 (marginal/conditional) N/groups AIC BIC

I – – – 0.000 0.001 0.009 0.106 2.434** 0.527*** 0.393** 0.155*** – – – – 20.462*** 2.348 0.007/0.680 2653/15 10189.59 10254.31

II 0.501*** 0.527*** 0.824*** – – – – – – – – – – – – 19.412 *** 2.376 0.008/0.688 2653/15 10154.89 10190.19

III – – – – – – – – – – – 1.336 0.479 0.567 2.895* 21.997*** 1.804 0.297/0.684 2653/15 10219.76 10260.94

IV 0.463*** 0.591*** 0.899*** 0.000 0.075 0.028 0.016 2.549*** 0.420*** 0.379** 0.253*** 0.975*** 0.229* 0.722*** 3.135*** 21.400*** 1.818 0.311/0.701 2653/15 10118.7 10224.61

***99%, **95%, *90%

approach with personality traits and the body signals itself. Lastly, we created one regression model containing all dependent variables. We find that a neutral arousal state and high arousal both have a positive impact on self-perceived creativity. A high pleasance state also has a positive impact on creativity. Comparing those results with the body signals, we find that the accelerometer average and variability seem to have a slightly weaker coefficient. However, going for a run in the morning has a greater impact on creative performance than the different mood-related variables.

4.4

Personality, Body Signals, and Creativity

As a second baseline, we chose the personality of the participants. Looking at the correlation between the different personality traits and the creative performance, we find that most of the personality traits seem to influence creativity strongly. The openness of participants is a strong indicator for creative individuals, and being agreeable seems to strongly mitigate the perceived creativity of an individual as does being conscientious. Extroversion has a weak impact on creativity.

196

M. Stein et al.

For the regression analysis, we find the different personality traits to mostly be insignificant in model III. However, if we look at the combined model IV, we find all personality traits to be significant; this indicates that personality traits alone are an insufficient predictor for perceived creative performance and that combining those variables with physiological measures improves the quality of the regression model.

4.5

Summing Up

We find the predictive power of body signals toward self-perceived creativity on a similar level as the predictive power of mood states and personality toward selfperceived creativity. The regression coefficients of the personality-related variables seem to be somewhat more predictive than the coefficients of the body signals, yet they are insignificant in a model that does not contain body signals. Therefore, our hypothesis seems to be confirmed: Body signals captured by a smartwatch correlate significantly with an individual’s self-perceived creative performance. Moreover, our findings are in line with prior research analyzing the relationship between body signals and creativity or performance. As introduced in Sect. 2, Pentland (2010) analyzed the performance of business pitches by using the honest signals as a predictor. They found a high level of activation (measured by increased movement) to be one of the predictors for a successful pitch. In addition, they found low consistency (measured by a high variability in movement) to mitigate success in that context. We were able to replicate these findings with self-perceived creativity as the dependent variable and different movement-related variables as well as the heart rate as independent variables. We found the honest signal activity, measured through movement intensity (acc. avg., VMC, Activity: walk, Activity: run), reflects self-perceived creativity. Low consistency on the other side, measured through movement variability (acc. var.), reflects low self-perceived creativity. We acknowledge that some of the correlations might not reflect causality. For example, moving more intensively does not necessary mean that one person is more creative in that moment. However, what our correlation and regression analysis does show is that there are patterns in the data that link body signals captured by a smartwatch to self-perceived creativity. Although the full extent of the relationship probably cannot be caught by a simple correlation or regression analysis, our work shows that this relationship is worth studying. Further, we argue that smartwatch-based body sensors are a viable tool for measuring different honest signals in social interaction because they remedy many of the limitations outlined in the first section. For instance, current methods to collect physiological data for creativity research are costly (1) and thus limit the sample size of such studies (2), and (3) past studies using physiological data to find influencers on creativity employ intrusive equipment in order to collect the body signals. Regarding limitation (1) and (2), we found that off-the-shelf smartwatches are accurate enough to find metrics that correlate with the creative performance of individuals. They therefore mitigate the necessity of costly equipment to measure

Show Me Your Moves: Analyzing Body Signals to Predict Creativity of. . .

197

body signals. Hospital grade heart rate monitors cost multiple thousands of dollars and are restricted to be used by only one person at a time (Gaskin et al. 2017). Our Pebble smartwatch in contrast can be obtained for only around $100. Moreover, this reduction in costs allows us to conduct studies with a bigger sample size. Gaskin et al. (2017) found that many prior studies analyzing the effects of physiological data on different outcome variables are severely limited by their sample size. They argue that the low sample size is partly due to the prohibitive cost of using equipment to collect such data. Utilizing a simple smartwatch overcomes this limitation and allows researchers to do studies with more participants. However, we also identify some problems with smartwatch-based physiological measures. When analyzing the heart rate measure taken by our smartwatch, we found that it cannot be considered a reliable predictor for creative performance. Although we found a significant correlation between these two variables, when used in a regression model, the heart beat variable was mostly uninfluential. Further research has to show whether this effect is caused by measurement errors of the smartwatch or by our way of preparing the dataset. Regarding limitation (3) we argue that smartwatches are less obtrusive than other devices that are usually employed to collect body signals. Professional instruments that collect body signals usually require extensive tethering of the individuals. Although sociometric badges already mitigate this problem, they still don’t extinguish it. Letting individuals wear a sociometric badge might still influence their natural behavior. Therefore, usage of sociometric badges might confound the actual measures taken during such a study. Especially if we try to identify behavioral patterns, a measuring device that influences those patterns might lead to skewed results. A smartwatch on the other side is as unobtrusive as a bracelet. Although we didn’t control for this effects in our study, Gaskin et al. (2017) did when conducting a similar study. They found that the vast majority of participants indicated that they were not influenced by the smartwatch they used.

5 Limitations and Future Work The major limitation of this study is the small sample size which constrains the validity of our findings with just 15 participants over 3 days. For example, participants only indicated pleasant mood states during the time of the study. Another issue with the sample size can be found for the variable Activity: run, which only occurred three times in our dataset. Moreover, all instances of this variable resulted from the same person going for a 45-min run in the morning. Further studies have to show whether there is general relationship between going for a run and selfperceived creativity. Yet another limitation is the metric by which creativity was measured. Selfevaluated creativity might not reflect the actual creativity during the group work process. Future studies should reevaluate the findings with more objective measures.

198

M. Stein et al.

Additionally, we are working on improving the employed smartwatch application to include more sensors to get a more holistic view of the participants’ body signals. We aim to replicate this study with an extended version of our body sensing smartwatch which will able to capture the speech energy of the participants as well as the physical proximity between the participants. Collecting those measures will enable us to not only analyze the body signals on the individual level but to also take the differences between the body signals of the different participants within a group into account. Despite the aforementioned limitations, we showed that body sensing smartwatches can be a viable tool to measure creativity and performance of knowledge workers in organizational settings. Thus, we encourage fellow researchers to employ similar methods of data collection to further deepen our understanding of how physiological measures can reflect creativity and other hard to measure personality traits.

References Abdel-Khalek, A. M. (2006). Measuring happiness with a single-item scale. Social Behavior and Personality: An International Journal, 34, 139–150. Adaman, J. E., & Blaney, P. H. (1995). The effects of musical mood induction on creativity. Journal of Creative Behaviour, 29, 95–108. Amabile, T. M. (1988). A model of creativity and innovation in organizations. Research in Organizational Behavior, 10, 123–167. Aral, S., Brynjolfsson, E., & Van Alstyne, M. (2012). Information, technology, and information worker productivity. Information Systems Research, 23, 849–867. Ashby, F. G., Isen, A. M., et al. (1999). A neuropsychological theory of positive affect and its influence on cognition. Psychological Review, 106, 529. Baas, M., De Dreu, C. K., & Nijstad, B. A. (2008). A meta-analysis of 25 years of mood-creativity research: Hedonic tone, activation, or regulatory focus? Psychological Bulletin, 134, 779. Banfield, R., Lombardo, C. T., & Wax, T. (2015). Design sprint: A practical guidebook for building great digital products. Newton, MA: O’Reilly Media. Barrick, M. R., & Mount, M. K. (1991). The big five personality dimensions and job performance: A meta-analysis. Personnel Psychology, 44, 1–26. Bell, R., & Lumsden, J. (1980). Test length and validity. Applied Psychological Measurement, 4, 165–170. Bond, M. H., Nakazato, H., & Shiraishi, D. (1975). Universality and distinctiveness in dimensions of Japanese person perception. Journal of Cross-Cultural Psychology, 6, 346–357. Bowers, M. T., Green, B. C., Hemme, F., & Chalip, L. (2014). Assessing the relationship between youth sport participation settings and creativity in adulthood. Creativity Research Journal, 26, 314–327. Budner, P., Eirich, J., & Gloor, P. A. (2017). “Making you happy makes me happy” - measuring individual mood with smartwatches. ArXiv171106134 Cs. Clapham, M. M. (2001). The effects of affect manipulation and information exposure on divergent thinking. Creativity Research Journal, 13, 335–350. Conley, J. J. (1985). Longitudinal stability of personality traits: A multitrait–multimethod– multioccasion analysis. Journal of Personality and Social Psychology, 49, 1266. Costa, P. T., & McCrae, R. R. (1989). NEO five-factor inventory (NEO-FFI). Odessa FL Psychol. Assess. Resour.

Show Me Your Moves: Analyzing Body Signals to Predict Creativity of. . .

199

Daggett, M., O’Brien, K., Hurley, M., & Hannon, D. (2017). Predicting team performance through human behavioral sensing and quantitative workflow instrumentation. In Advances in human factors and system interactions (pp. 245–258). Heidelberg: Springer. Digman, J. M. (1990). Personality structure: Emergence of the five-factor model. Annual Review of Psychology, 41, 417–440. Digman, J. M., & Inouye, J. (1986). Further specification of the five robust factors of personality. Journal of Personality and Social Psychology, 50, 116. Gaskin, J., Jenkins, J., Meservy, T., Steffen, J., & Payne, K. (2017). Using wearable devices for non-invasive, inexpensive physiological data collection. In Hawaii international conference on system sciences 2017 HICSS-50. Glover, J. A., Ronning, R. R., & Reynolds, C. (2013). Handbook of creativity. Cham: Springer. Göritz, A. S., & Moser, K. (2003). Mood and flexibility in categorization: A conceptual replication. Perceptual and Motor Skills, 97, 107–119. Hills, P., & Argyle, M. (2002). The Oxford happiness questionnaire: A compact scale for the measurement of psychological well-being. Personality and Individual Differences, 33, 1073–1082. Isen, A. M., & Baron, R. A. (1991). Positive affect as a factor in organizational-behavior. Research in Organizational Behavior, 13, 1–53. Jausovec, N., & Bakracevic, K. (1995). What can heart rate tell us about the creative process? Creativity Research Journal, 8, 11–24. Knez, I. (1995). Effects of indoor lighting on mood and cognition. Journal of Environmental Psychology, 15, 39–51. Léger, P.-M., Davis, F. D., Cronan, T. P., & Perret, J. (2014). Neurophysiological correlates of cognitive absorption in an enactive training context. Computers in Human Behavior, 34, 273–283. Lyubomirsky, S., King, L., & Diener, E. (2005). The benefits of frequent positive affect: Does happiness lead to success? Worcester, MA: American Psychological Association. McCrae, R. R., & Costa, P. T. (2004). A contemplated revision of the NEO five-factor inventory. Personality and Individual Differences, 36, 587–596. Mikulincer, M., Kedem, P., & Paz, D. (1990). Anxiety and categorization—1. The structure and boundaries of mental categories. Personality and Individual Differences, 11, 805–814. Minnery, B. S., & Fine, M. S. (2009). FEATURE neuroscience and the future of human-computer interaction. Interactions, 16, 70–75. Mumford, M. D. (2003). Where have we been, where are we going? Taking stock in creativity research. Creativity Research Journal, 15, 107–120. Norman, W. T. (1963). Toward an adequate taxonomy of personality attributes: Replicated factor structure in peer nomination personality ratings. Journal of Abnormal and Social Psychology, 66, 574. Pentland, A. (2010). Honest signals: How they shape our world. Cambridge, MA: MIT press. Posner, J., Russell, J. A., & Peterson, B. S. (2005). The circumplex model of affect: An integrative approach to affective neuroscience, cognitive development, and psychopathology. Development and Psychopathology, 17, 715–734. https://doi.org/10.1017/S0954579405050340 Robins, R. W., Fraley, R. C., Roberts, B. W., & Trzesniewski, K. H. (2001). A longitudinal study of personality change in young adulthood. Journal of Personality, 69, 617–640. Russell, J. A., & Pratt, G. (1980). A description of the affective quality attributed to environments. Journal of Personality and Social Psychology, 38, 311. Somech, A. (2006). The effects of leadership style and team process on performance and innovation in functionally heterogeneous teams. Journal of Management, 32, 132–157. Sternberg, R. J. (1999). Handbook of creativity. Cambridge: Cambridge University Press. Stockham, N. (2016). Opening the black box: Publishing Pebble’s activity-tracking algorithms. Medium.

200

M. Stein et al.

Sung, S. Y., & Choi, J. N. (2009). Do big five personality factors affect individual creativity? The moderating role of extrinsic motivation. Social Behavior and Personality: An International Journal, 37, 941–956. SuzanneK. Vosburg, G. K. (1997). ‘Paradoxical’ mood effects on creative problem-solving. Cognition & Emotion, 11, 151–170. Thayer, J. F., Ahs, F., Fredrikson, M., Sollers, J. J., & Wager, T. D. (2012). A meta-analysis of heart rate variability and neuroimaging studies: Implications for heart rate variability as a marker of stress and health. Neuroscience and Biobehavioral Reviews, 36, 747–756. Verhaeghen, P., Joorman, J., & Khan, R. (2005). Why we sing the blues: The relation between selfreflective rumination, mood, and creativity. Emotion, 5, 226. Vosburg, S. K. (1998). The effects of positive and negative mood on divergent-thinking performance. Creativity Research Journal, 11, 165–172. Wagner, J., Kim, J., & André, E. (2005). From physiological signals to emotions: Implementing and comparing selected methods for feature extraction and classification. In IEEE international conference on Multimedia and expo, 2005, ICME 2005 (pp. 940–943). IEEE. Yamamoto, K. (1963). Relationships between creative thinking abilities of teachers and achievement and adjustment of pupils. The Journal of Experimental Education, 32, 3–25.

Promoting and Supporting Biodiversity Conservation Activities with the Pattern Language Approach: A Pattern Language for Collaborative Activities for Biodiversity Conservation Arisa Kamada, Konomi Munakata, Mahito Kamada, Tomohiro Ichinose, and Takashi Iba Abstract In this paper, we present a pattern language comprising 23 patterns for collaborative activities for biodiversity conservation. These patterns have been created based on interviews with people working in Kitahiroshima which is known in Japan for its successful model in biodiversity conservation. We extracted the knowledge from practice: how they have been constructing Collaborative Innovation Networks. We expect this outcome to be applied to various areas and support the next generations of the communities for biodiversity conservation.

1 Introduction In Japan as well as other developed countries, the quality and quantity of the ecological services have been dramatically lost due to the loss of biodiversity. Activities for conservation and the suitable usage of the ecosystem have been therefore strongly recommended and promoted. In some areas, such activities have been successfully working and producing good results, whereas others have been struggling with problems.

A. Kamada (*) Graduate School of Media and Governance, Keio University, Tokyo, Japan e-mail: [email protected] K. Munakata · T. Iba Faculty of Policy Management, Keio University, Fjisawa, Kanagawa, Japan M. Kamada Graduate School of Technology, Industrial and Science, Tokushima University, Tokushima, Japan T. Ichinose Faculty of Environment and Information Studies, Keio University, Fjisawa, Kanagawa, Japan © Springer Nature Switzerland AG 2019 Y. Song et al. (eds.), Collaborative Innovation Networks, Studies on Entrepreneurship, Structural Change and Industrial Dynamics, https://doi.org/10.1007/978-3-030-17238-1_12

201

202

A. Kamada et al.

2 Background The actions needed in conserving biodiversity largely differ depending on the situation of each region. Therefore, a solution proved to be successful in one specific case cannot be always successful in another location. Furthermore, successful cases often have a key person whose contribution plays a crucial role. In such a case, it is often assumed that the success of that project was brought about solely because of the key person’s individual skills. In addition, a collaboration of different stakeholders, not only the experts of a topic, is necessary in order to secure enough personnel for biodiversity conservation in each region. However, it is not easy for non-experts to recognize the problem and truly understand the importance of preserving biodiversity and related activities, because the impact of problems due to the loss of biodiversity appears in a long run and cannot be seen on a day-to-day basis. Meanwhile, in Kitahiroshima, Hiroshima Prefecture, an initiative to preserve biodiversity through a new social system is taking place and represented as a restructuring model in Japan. They created a system that can be operated by community members, whereby a better relationship between human and nature has been constructed. Here, the contribution of Mr.S who got to the town as a curator became a Creator, thereby gradually forming the Collaborative Innovation Network. What was done in these practices that are known as successful models in Japan? In particular, how could different stakeholders form good relations which enabled the projects to work well? This research aims to generalize the essential patterns gathered through interviews with key persons in the model case and make them able to be applied to other locations using the methodology of pattern language. The pattern language approach would lessen the focus on individuals’ abilities and actively join more people to the activities toward improving the situation. Also, it is expected for each project to understand the essential patterns needed and apply proven practices in their own contexts.

3 Method: Pattern Language Pattern language is a methodology of extracting and sharing the knowledge of practice from individual cases and experiences. It works as a common language among different people from different areas and therefore enables collaborative work. This methodology, invented in the domain of architecture (Alexander et al. 1977), has been applied to software design, as well as various human action topics. Recent topics of application include education (Bergin et al. 2012), team reformation (Manns and Rising 2005), presentation (Iba and Laboratory 2011), and welfare (Iba et al. 2015). In this research, we applied this method to the domain of promoting and supporting biodiversity conservation activities.

Promoting and Supporting Biodiversity Conservation Activities with the. . .

3.1

203

Format of Each Pattern

The pattern language method captures the practical knowledge of a certain topic in the unit of “patterns.” A pattern language is the result of loosely structured patterns, each with their own “name,” into a whole and comprehensive form. Each pattern is written in a specific format that summarizes the context, problem, solution, and consequence. In other words, each pattern expresses the knowledge of a certain context related to the frequency of a specific kind of problem. The pattern indicates what is a good solution for that problem and what will happen when it is successfully implemented (Fig. 1). In this way, the pattern language makes it possible to share the essential patterns needed for something to work well. In addition, because each pattern has a “name,” which works as common vocabulary, stakeholders across different projects and sectors can discuss, learn from others, and collaboratively work together.

3.2

Creating Process

In order to create a pattern language, it is first necessary to extract knowledge of practice through the investigations of successful models and by interviewing people (Phase 1: Pattern Mining). In the next phase, a full description of the patterns (context, problem, solution, and consequence) is written based on the essence of the patterns (Phase 2: Pattern Writing). This description is then subject to a detailed collaborative revision process. In the final phase, a Pattern Name and Pattern

Fig. 1 A format of pattern language

204

A. Kamada et al.

Illustration are added to express the essence of patterns symbolically (Phase 3: Pattern Symbolizing). During these processes, certain phases are revisited/repeated as the language develops, as necessary.

4 Study Area: Kitahiroshima Formerly in Japan, biodiversity used to be sustained by cyclic use of forest materials. Especially, early successional species such as grassland plants or animals or insects were targeted. However, globalization has caused imported fossil energy to be the main source instead of wood energy. Forests were neglected as a result of its lost economical value, and consequently, habitat for grassland species has been lost due to progressive succession. Also, the relationship between human and nature has been lost in Japan, and it is necessary to restructure this relationship. In Kitahiroshima, the NPO-led initiative brings together forest owners, local stores, hotels, and consumers to promote the usage of wood as an energy source. The initiative contributes to the local economy by circulating firewood used as local currency and also aims to regenerate ecosystems through the usage of forests (Fig. 2). The system motivates community member to participate in the activity (Kamada 2018). In this area, grassland species are recovering gradually.

Fig. 2 The system in Kitahiroshima (adopted from Kamada 2018)

Promoting and Supporting Biodiversity Conservation Activities with the. . .

205

Onishi (2015) has showed the process and the stakeholders’ network of this activity. As a result, it became clear that the network was especially expanded by Mr.S, the member of the NPO. In this research, therefore, the focused is on “how” he has actually realized the great collaboration as a coordinator.

5 Investigation Process The pattern language was created through the following processes.

5.1

Interviews

Using the method of Iba and Yoder (2014), interviews were used to gather knowledge on how the project was conducted, including member S (key person/NPO and city museum member), K (key person/NPO member), KN (NPO member), BB (government staff), and CC and DD (both elementary school staff). The contents of all interviews were recorded and key points written down individually on sticky notes. In this phase, we gathered mainly parts for the solution and consequence (Fig. 3).

5.2

Clustering

The data extracted from the interviews were clustered and structured, using the KJ Method (Kawakita 1967) (Fig. 4). Fig. 3 The gathering part

206

A. Kamada et al.

Fig. 4 Clustering with stick notes

Fig. 5 Writing process

5.3

Writing

A full description of the patterns (context, problem, solution, and consequence) is written down based on the essence of the patterns. We assumed context and problem from the solution and consequence (Fig. 5).

Promoting and Supporting Biodiversity Conservation Activities with the. . .

5.4

207

Reviewing

Prototypes were developed through the discussion with stakeholders.

6 Result As a result, 23 patterns were extracted and categorized into four groups. Patterns in the category “start up the project” show practical knowledge for the beginning of the activities. This category contains seven patterns. Patterns in the category “warm up to the project” share practical knowledge for getting the project on the right track. This category contains three patterns (Fig. 6). Patterns in the category “respect the collaborators’ context” introduce practical knowledge for making good connection with collaborators. This category contains seven patterns. Lastly, patterns in the category “move things forward” are designed to move the activities forward. This category contains six patterns (Fig. 7). The contents of all patterns can be seen in the appendix.

Fig. 6 Pattern Names1

Fig. 7 Pattern Names2

208

A. Kamada et al.

7 Discussion A notable point discovered in this research is that biodiversity conservation activities in Kitahiroshima are designed as a part of the region’s effort in “community building centered on natural conservation.” That is, they propose the idea that community efforts to protect its nature lead to the vitalization of local communities. It is not enough to advocate the importance of wild animals or plants but also the need to inform the usability for the community. Furthermore, one of the keys to maintain such collaboration is to have some intentional “slack,” which allows even people without knowledge in nature and biology to “casually” participate. It is difficult to attract the local people’s interest when the project is positioned as “biodiversity conservation.” Therefore, even if it seems indirect, it is crucial to position it as a relevant issue that must be tackled through the collaboration of the region’s various stakeholders.

8 Expect Outcomes The pattern language approach makes it possible to share the essential patterns needed for something to work well. Therefore, it lessens the focus on individuals’ abilities and enables each project to apply proven practices in their own contexts. Because each pattern has a “name,” the patterns themselves function as common vocabulary. Stakeholders across different projects and sectors can discuss, give advice, ask common vocabulary, questions, and learn from one another. Furthermore, by visualizing the key persons’ implicit skills, it is possible to define his/her role and the significance they play. This creates a better possibility of the project being able to acquire the proper budget needed to maintain and run the project. We are still searching for more words to include this pattern language in the future that can help people to promote biodiversity conservation activities, and we will continue our research in order to update the language.

9 Future Work In order to elaborate on the pattern language, we will go through more investigations in different areas in Japan. We will apply this pattern language to other areas in Japan to support and evaluate the biodiversity conservation activities. In the first year, we have conducted an investigation in Kitahiroshima in Hiroshima Prefecture. Additional investigations are now planned in the following three areas.

Promoting and Supporting Biodiversity Conservation Activities with the. . .

9.1

209

Fukutsu City in Fukuoka Prefecture

This city has been developed as a commuter belt with a growing population since 2010. The forest, however, has been damaged. A part of this city is a specified area named “Genkai National Park,” and very important for the story on Satoyama habitat, but has been facing environmental problems such as the lack of control and the invasion of bamboo. Therefore, a laboratory of environmental design in Kyushu University has taken initiative to improve the situation by collaborating with the citizens, local groups for environmental conservation, and high schools (Ehime Prefectural Fisheries High School and Kanagawa Prefectural Koryo High School), learning the ecological system, and sharing information with others (Ito et al. 2017).

9.2

Aya in Miyazaki Prefecture

In September 2011, the elephant tree forest in this town was recommended for UNESCO Eco Park, and it was officially registered in July 2012. Since then activities have been promoted to create a mechanism for residents to take part in the community development, including the preservation of the local living forest. Efforts were made as a public-private partnership to thin our artificial forests and restore the natural forests. Many achievements have been drawing attention, as it has enabled the sustainable coexistence of people and nature (Manns and Rising 2005).

9.3

Tokushima Prefecture

In the formulation of the Biodiversity Regional Strategy of Tokushima Prefecture (Tokushima Prefecture 2014), the “Biodiversity Tokushima Conference” which was voluntarily organized by 18 civil social groups, researchers in Tokushima Prefecture played a major role (Kamata 2013). The conference organizes town meetings in various parts of the prefecture, in which the members collect and consolidate opinions of residents as facilitators and link various entities such as prefectures, experts, committees, and companies/business operators to promote the collaborative activities. Acknowledgments We first acknowledge interviewees to share the knowledge that is very helpful to others. Additionally, we would like to thank those who helped us write this paper, including Yang Song, the reviewers, and readers who gave feedback.

210

A. Kamada et al.

Appendix Pattern name Context Start up the project Intriguing You need to find people proposal who are willing to work together to move the project forward Personal enticement

You want to call for volunteers who are willing to implement the project together

Passion stock

You got volunteering members in your project

Risk forecast

You are working with people gathered as volunteers

Room for progress

You received opinions and requests from volunteers

Attraction as participant

You want to carry on the project, having new volunteering members continually recruited You want many people to engage in the activity on an ongoing basis

Face-to-face connections

Problem

Solution

It is hard to get new members, if you just call for somebody’s participation in front of many people Inviting a wide range of people does not attract enough number of people

Make a proposal for each one of people you want to invite, and tell them what exactly you want to do with them Send letters or e-mails directly to the people who might be interested in your activity or would never feel uncomfortable with the invitation Ensure that every members can sustain their motivation as a volunteer

If their motivation to commit to the project diminishes, it is difficult to keep the project on track Some people may not follow the rules or just take pictures and do not concentrate on the activity, but pointing them out during the activity may ruin the atmosphere You may have unintentionally drawn a clear line between you and volunteers as managers and participants and therefore cannot share the important roles and information with them If you explain the attractive part of the project by yourself, it may sound like a lie If there are few opportunities to meet up, their motivation easily diminishes

Inform the important rules beforehand to those who have decided to participate in the activity

Give them appropriate roles and tasks, depending on what they want to do, and share the purpose firmly

Pick up words from the actual participants

Keep communicating with participants so that they can easily think of your face (continued)

Promoting and Supporting Biodiversity Conservation Activities with the. . . Pattern name Context Warm up to the project Easy entrance You want the local people get interested in the nature and the environmental conservation

Problem

Solution

Even if you hold a symposium or events, most participants are from outside, not the local people

Allow the “laid-back participation,” by inviting guests who are not directly related to the environmental activities and making it easy to jump in Have the discussions with eating and drinking, or set a tea time in between to make the workplace more comfortable Ask questions to get constructive ideas for the next time

Comfortable workplace

You want the local people get interested in the nature and the environmental conservation

When thinking about the nature, environmental conservation, and the future, the atmosphere is likely to get too serious

Progressinducing request

You want to hear the opinions and needs from people who participated in the local event

Even if you have prepared a questionnaire, you may not be able to get honest voices from them

Respect the collaborators’ context Start by You want to collaborajoining tively work with people from different sectors in that region Passion You want to collaborawithin tively work with people chatting from different sectors in that region Individual proposals

You want to collaboratively work with people from different sectors in that region

Familiar words

You want to collaboratively work with people from different sectors in that region

Overlap of words

You want to work firmly with the administration

211

If various organizations stick with their own context, the collaboration will never work It is difficult to fully understand what people need, and thus it is difficult to always make appropriate suggestions It is difficult for people to understand and imagine the actual activity taking place, when they just hear the keywords such as “environmental conservation” and “natural learning” If the concepts and keywords of the project are not familiar with the local people, they will not get empathy to take actions Due to the nature of administrative works, it is uneasy to get flexible responses and supports, and therefore conflicts often occur

Join the work first

Assemble the needs and complains from the daily conversations

Make proposals according to the personal experience, taking account of their living place and belonging organizations

Express your goal image, using the dialects or distinctive phrase in that region

To make the work easier for administrators, make a proposal with the goals and keywords described in the administrative official documents (continued)

212 Pattern name Contextdriven proposals

Merit for partners

A. Kamada et al. Context You want to educate children in the region about the local nature and environmental conservation You have been driving the project with people from other sectors in the region

Move things forward Visible You are requesting resolution cooperation from other stakeholders for starting a new project Explanation chain

You are requesting cooperation from other stakeholders for starting a new project

Trigger for change

Various opposing voices were raised by starting the project as it is a different approach from previous mechanisms and practices You are collaboratively working with people from different sectors in that region

Collaborative problemsetting

Shared vision

You want to apply the successful method which has been working well in other areas in your own region

Speed booster

You went to visit other areas with local people

Problem In the framework of existing curriculum and courses, it is impossible to provide attractive classes for children Even if you propose ideas and opinions to make the situation better, it is difficult to be convinced

Solution Grasp the flow of the education and suggest ideas accordingly

It is often difficult to gain understanding of others, and it takes time to start the project more than necessary You may have no idea how to start explaining the problematic situation to persuade them

Make up your mind to implement the project even all by yourself and show it to them

Trying to explain and persuading everyone, you will run out of energy and time

The project is likely to get stuck, when the problem is differently recognized among stakeholders Just learning the mechanism and mimicking it does not work

It is likely to end up with just saying “it was a great model” and bring back nothing to improve your region

Make the proposal from the others’ points of view to strengthen its merit for them

First, go talk to the people who you can access easily, and explain your situation persuasively one after another Do not be swayed by everyone’s opinions; think about where to put your energy

Conduct a field work, figure out, and state the problem together

Visit to see the successful model with the administrative communities and the works that are working together, and think about what you can do in your area Think about how you can realize what you have learned, and put it into words as soon as possible

Promoting and Supporting Biodiversity Conservation Activities with the. . .

213

References Alexander, C., Ishikawa, S., Silverstein, M., Jacobson, M., Fiksdahl-King, I., & Angel, S. (1977). A pattern language: Towns, buildings, construction. New York: Oxford University Press. Bergin, J., Eckstein, J., & Pedagogical Patterns Editorial Board. (2012). Pedagogical patterns: Advice for educators. Scotts Valley, CA: Createspace. Iba, T., & Iba Laboratory. (2011). Presentation patterns: A pattern language for creative presentations. Tokyo: Keio University Press. Iba, T., Okada, M., & Dementia Friendly Japan Initiative, & Iba Laboratory. (2015). Words for a journey: The art of being with dementia. Yokohama: CreativeShift Lab. Iba, T., & Yoder, J. (2014, November). Mining interview patterns: Patterns for effectively obtaining seeds of patterns. In SugarloafPLoP’14 proceedings of the 10th Latin American conference on pattern languages of programs. Ito, K., Sudo, A., Hanada, Y., Nakamatsu, T., Ikejiri, E., & Lin, S. Y. (2017). Kankyo Kihon Keikaku Seibutstu Tayousei Chiiki Senryaku to Rendoushita Shiminsanka niyoru Shinrin KanriFukutsushi niokeru Project Management to Sono Kadai [Forest management through citizen participation linked with the basic environmental plan and biodiversity regional strategy: Project management in Fukutsu city and its problem] (in Japanese). In The 128th Japanese Forest Society, Session ID, A12. Kamada, M. (2018). Satoyama landscape of Japan-past, present and future. In S.-K. Hong & N. Nakagoshi (Eds.), Landscape ecology for sustainable society (pp. 87–109). Cham: Springer. Kamata, M. (2013). Seibutstu Tayousei Chiiki Senryaku no Sakutei to Suishin niokeru Kyoudou [Collaboration in the formulation and promotion of biodiversity regional strategy], (in Japanese). Japanese institute of landscape. Architecture, 77(2), 95–98. Kawakita, J. (1967). Hassouho: Sozosei Kaihatsu notameni [Abuduction method: For development of creativity] (in Japanese). Chuokoronsha. Manns, M. L., & Rising, L. (2005). Fearless change: Patterns for introducing new ideas. Boston: Addison-Wesley. Onishi, M. (2015). Kuukan no Rireki wo Ikashita Kyodo process no design to management: Hiroshimaken Kitahiroshimamatchi no Seitaikeihozen Katsudo wo Jireitoshite [Design and management of collaborative processes with the historical usage of space: Case study on biodiversity conservation activities in Kitahiroshima Towan in Hiroshima prefecture] (in Japanese). Tokushima: Institute of Technology and Science, the Tokushima University.

“Twelve-Tone Music Reloaded”: 12 Lessons in Rotating Leadership and Organizational Development from Jazz Daniel C. Schmid and Peter A. Gloor

Abstract This paper illustrates the core principle of COINs (collaborative innovation network) of rotating leadership by the example of Jazz musicians, who take turns grooving together. These musicians are exemplars of team members seamlessly transferring the leadership role from one to the other, leading to a “flow” experience of superb quality for their audience. As we show, so-called honest signals from Jazz can play a key role for organizational development to create an “organizational groove.”

1 Introduction In the past 25 years, Jazz has enjoyed increasing popularity as a metaphor for collaborative organizational development, functioning team roles, and leadership structures (Weick 1999; Kamoche et al. 2003; Burow 2004; De Pree 2008; Barrett 2012). Based on the revolutionary field theory of Kurt Lewin (Lewin 1997), Peter Senge’s groundbreaking study of the learning organization (Senge 2006), and Ed Schein’s decades of work about organizational culture (Schein and Schein 2017), the success of the metaphor seems not only to be a counter-movement toward top-down leadership approaches. It’s also been motivated by the fact that Jazz is a social phenomenon in which improvisation is much more salient (Kamoche et al. 2003). Recently, Jazz improvisation has become a part of the “Holy Grail” in organizational development, above all under so-called VUCA (volatility, uncertainty, complexity, and ambiguity) conditions (Johansen 2012). Effective leadership of the future will be based on relationship building, understanding complex group work, and diverse

D. C. Schmid (*) HWZ University of Applied Sciences in Business Administration Zurich, Zurich, Switzerland e-mail: [email protected] P. A. Gloor MIT Sloan Center for Collective Intelligence, Cambridge, MA, USA e-mail: [email protected] © Springer Nature Switzerland AG 2019 Y. Song et al. (eds.), Collaborative Innovation Networks, Studies on Entrepreneurship, Structural Change and Industrial Dynamics, https://doi.org/10.1007/978-3-030-17238-1_13

215

216

D. C. Schmid and P. A. Gloor

workforces. For future leaders, it will be key to gain a deeper understanding of the constantly evolving complexities of interpersonal, group, and even intergroup relationships (Schein and Schein 2018). Rotating leadership is one of the key tenets of collaborative innovation networks (COINs). COINs are small groups of intrinsically motivated people who get together to create something radically new. While earlier research on COINs demonstrated the need for strong leaders (Nemoto et al. 2011), a COIN with a single leader stagnates and loses its innovative spark (Kidane and Gloor 2007). High-functioning COINs have multiple leaders who take turns providing leadership according to their skills and strengths. They are as willing to step back when somebody else is more qualified for a task, as they are willing to step up when they see the opportunity to contribute their expertise and capabilities. Cooperation among Jazz musicians is an exemplar of rotating leadership (Gloor et al. 2013). In earlier work this has been researched using sociometric badges, bodyworn devices that record the body signals of the musicians. In this paper we first discuss key principles of rotating leadership at the example of two ragtime pianists, Tom Brier (www.last.fm/music/Tom+Brier/+wiki) and Adam Swanson (www. adamgswanson.com), complementing them with five more general lessons for organizational design derived from Jazz musicians. Tom Brier and Adam Swanson are well known in the field, having won different prizes and having performed at many festivals. Both of them also started as child prodigies, discovering their calling as professional ragtime pianists early on in their lives. Over the years both had been performing and competing at different festivals across the USA. In this paper we would like to focus on two YouTube videos recorded in 2008 and 2010, where Tom and Adam out of the blue began playing the “chopsticks” rag together. The first time this happened was at the tenth Sutter Creek Ragtime Festival (www.youtube.com/watch?v¼H-8bSZa4Fjg) and the second time was at the West Coast Ragtime Festival (www.youtube.com/watch?v¼ku2r8HdKong). The first time, in 2008, Adam had been playing the first few bars of “chopsticks” as a joke, not knowing how Tom would respond to it. Also, at that time, Adam did not know the full “chopsticks” rag after the first few bars. Tom took up the challenge and started playing; Adam had to pick up the tune and learn it on the fly. The second time, in 2010, Adam was well prepared, and Tom and Adam started grooving for 9 min, as recorded in the video. In the 2008 event, Adam turned “chopsticks” first into the “Tim Baileys” rag and then, both in 2008 and 2010, into the “Tiger Rag.” Along the way, Tom and Adam switched keys on each other, thus creating brief periods of dissonance, before getting back into synch. In their mutual interplay, Tom and Adam wonderfully illustrate the key principles of rotating leadership: 1. 2. 3. 4. 5. 6.

They are seamlessly passing control from one to the other. Whoever has a creative idea takes the lead, and the other follows. They are masters of their profession. The one who knows less learns from the master. They play and synchronize by ear. Competitive collaboration leads to the perfect product.

“Twelve-Tone Music Reloaded”: 12 Lessons in Rotating. . .

217

7. They do it for the fun of it! These seven key lessons in rotating leadership from the two ragtime pianists, Tom Brier and Adam Swanson, can be transferred into the broader organizational perspective. The main challenge if we want to consider Jazz as a possible metaphor for leadership development in organizations is that many of the decision-makers might have never been on stage and probably do not know about stage fright, lack of technical infrastructure, or the worst, bad vibes in a band. One of the authors has worked professionally for several years with international Jazz musicians such as Dave Brubeck, Ray Brown, Kenny Barron, Jim Hall, and many more. Based on this experience, the five following factors had the greatest impact on him as a Jazz aficionado before and behind the curtain: 8. 9. 10. 11. 12.

Be prepared. Learn to listen. Let it flow. Reduce to the max. Be modest.

Above all, these five key principles are completely analog and can be identified as “honest signals” (Gloor et al. 2013). So, in order to create a rotating leadership environment, these success factors can be aligned directly with the seven key principles we have learned before from Tom and Adam while playing together “Take five”! Let’s now look at each of these 12 lessons in detail, starting with the insights gained from observing Tom and Adam: 1. They Are Seamlessly Passing Control from One to the Other In each of the two improvisation sessions, Tom and Adam are switching the lead at least half a dozen times. Thanks to this rotation in leadership, creativity is flourishing, with unpredictable, but highly enjoyable and stimulating results. Initially Adam plays a few bars of “chopsticks,” which Tom then picks up and continues by making up a trio section on the spot. This is similar to real-world COINs, where leaders easily take turns in the leadership role, leading to an unpredictable but high-quality end product. 2. Whoever Has a Creative Idea Takes the Lead, and the Other Follows Initially Adam challenges Tom to play with him a more complex tune “Car-Barlick Acid.” Tom accepts, but when Adam starts playing “chopsticks” as a joke, Tom takes up this joke and keeps on playing “chopsticks,” improvising and adding new sections along the way until Adam changes their melody to “Bill Bailey,” flexibly picked up by Tom. Adam finally leads over into “Tiger Rag.” This again corresponds to real-world COINs where the leaders are in perfect synch, with the one who knows the tune best taking the lead and the rest of the team following along until another team member has a better idea, which is then adapted by the rest of the team.

218

D. C. Schmid and P. A. Gloor

3. They Are Masters of Their Profession Adam and Tom were both child prodigies who started at a young age playing ragtime. Playing rag is rarely a profession but always passion. In these two videos, the two artists are participating at a festival, but they still use the opportunity of having a break to play together even more. These two sessions are completely unplanned and unrehearsed; nevertheless they lead to a stunning performance, because the two masters speak the same (musical) language and are able to communicate through their music. They both have huge amounts of talent and passion for their art, which allows them to coordinate effortlessly and to easily switch roles. This is the same for any other COIN, where the one who is best qualified for a task becomes the leader until somebody else comes along who is better qualified. 4. The One Who Knows Less Learns on the Fly from the Master In their jamming session, the lead goes from Adam, who challenges Tom with a few bars of “chopsticks” to their musical competition, to Tom, who knows the tune and takes the lead, training Adam on the fly. Adam quickly masters it so well that he now can take the lead to successfully introduce “Bill Bailey,” bringing Tom along, who then steps up leading the way, until Adam again jumps ahead by switching over to the “tiger rag.” This means that COINs are also learning networks, where more junior COIN members are constantly trained by the master. 5. They Play and Synchronize by Ear In their communication and interplay, the “media is the message.” Adam and Tom speak the same musical language, even more, they breathe it, and their brains are wired in the same way. Both have been infected by the “ragtime virus” when they were 5 years old, and since then each has played for hours every day. They have the basic rhythms and melodies in their DNA and can play the “maple leaf rag” in their sleep. Cooperation and playing together comes as natural as breathing to them; they do not need conscious effort to collaborate; rather this happens in their subconsciousness. 6. Competitive Collaboration Leads to a Perfect Product If each of the players had jammed and improvised alone, never would an end product of this quality have come out. This session lives from the musical tension and creativity of these two top musicians, who nurture each other’s creativity in playful competition, thriving in creativity and trying to beat each other in new ideas, leading to perfect collaboration. This is competitive collaboration at its best! 7. They Do It for the Fun of It! Adam and Tom enjoy playing together so much that they utilize the break in their concert to play even more. Grooving and jamming for them is not work, but sheer pleasure that energizes them. COIN members join their group not for money, but because they are intrinsically motivated to work together toward their shared goal and develop a superior product. The remaining five principles resume general lessons for teamwork from Jazz, based on the personal experience of one of the authors.

“Twelve-Tone Music Reloaded”: 12 Lessons in Rotating. . .

219

8. Be Prepared None of the professional musicians we have ever met walked aimlessly to the stage. The set lists were always perfectly prepared, and, based on this fixed frame, flowing transitions, announcements, and improvised organizations were possible: preparation is everything, in Jazz as well as in leadership and management. For large organizations, which have to consider enormous scaling and scope effects, this means that the “set list” of their intended transformation has to be exactly designed in advance. The success of the “Real Book,” the Bible of Jazz notation, is based on the fact that it’s the perfect guideline and universally understandable. If Jazz musicians all over the world play or jam together, the “Real Book” is the “Lingua Franca of Jazz” and available in different keys (C/F/Bb/Eb) in order to avoid transpose misunderstandings among the different sections in the group (rhythm section, horns, electronic devices, etc.). 9. Learn to Listen For novices, Jazz mostly seems to be unstructured and a bit messy, in order to see through its sometime complex structures. Already at the early stage of Jazz, the great German philosopher Theodor W. Adorno, who escaped Nazi Germany toward England and later to the USA, refused to accept Jazz as art and claimed that it is at best “the everyday music of the upper class” (Adorno 1936). After WWII, JoachimErnst Berendt, at the time a young journalist who later should become the spiritus rector of international Jazz research, criticized Adorno in 1953. While Berendt considered American Jazz as the most original musical achievement of the twentieth century and emphasizes the high demands of the complex structures of Jazz based on improvisation, Adorno sees Jazz as a form of modern hit and accuses it of conventionality and conformism. Berendt, at the time, put it that way: “Adorno says that Jazz is ‘something in which there is nothing to understand but the rules of the game’. Well, at least that’s what it means to know what you’re talking about when you say Jazz” (Berendt and Adorno 1953). Once these structures are defined, it is clear how subjects are taken up and freely exchanged by the musicians. This requires, as in professional practice, to listen carefully and to ask if in doubt, because he who listens is leading, especially in business, as we learn from Otto Scharmer: “When you become a director, you will be confronted over and over with one thing: ‘Listen. Listen and learn to listen’” (Scharmer 2016, 2018). 10. Let It Flow According to Butterfield the microcosmos of Jazz opens its secrets by understanding the variation of the eight notes which enables a Jazz Band to “Swing”: “Varying the Beat-Upbeat Ratio (BUR)—i.e., moving between straight and swing eighth notes—enables Jazz musicians to manipulate the flow of motional energy across a phrase in systematic ways in conjunction with other melodic processes” (Butterfield 2011). In other words, the interaction and interplay of a Jazz group always bases on a common ground that has to be created first. Only by creating this secure base, improvising in Jazz will be possible at all, and recent analyses reveal the crucial importance of evaluating improvisational options in light of a field of interactive possibilities (Hodson 2007; Butterfield 2011). Once the audience is won, the

220

D. C. Schmid and P. A. Gloor

musicians are able to leave the security zone and venture into unknown musical spheres of an ordinary Jazz standard tune. Only those who are breaking the comfort zone are open to new experiences! 11. Reduce to the Max The most surprising fact in Jazz for one of the authors was that it is not important what is played, but what is NOT played. As in management, tempo, rhythm, and improvisation are the key factors for any successful corporate development. Simplicity is one of the main secrets in creating groove in Jazz. Some of the best examples ever recorded are “One Note Samba (Samba De Uma Nota So)” by Antonio Carlos Jobim or “All Blues,” one of the outstanding tunes in Miles Davis’ epochal oeuvre “Kind of Blue” (both released in 1959). Bill Evans, a Jazz giant himself and at the time pianist in Miles’ group, describes Davis’s ideas in the liner notes as “exquisite in their simplicity.” Without any rehearsal, and within only 2 days, “Kind of Blue” was recorded: the bestselling Jazz Album ever. Buster Williams, a renowned bass player who owned the biggest flight case one of the authors has ever seen in his life, described the magic of playing with Miles Davis as follows: “With Miles, it would get to the point where we followed the music rather than the music following us. We just followed the music wherever it wanted to go. We would start with a tune, but the way we played it, the music just naturally evolved.” (Williams, quoted by Barrett 2011). 12. Be Modest In all these years, the most successful musicians one of the authors has ever met were the most modest of them all. He is remembering, for example, private Japanese lessons with the legendary bass player Ray Brown or talking history with guitarist Jim Hall. Both were also excellent educators who could pass on their knowledge to the next generation: sharing is the magic power of rotating leadership, both in Jazz as in management. As we learn from Berendt, who later became a great music ethnographer, sharing and listening are highly connected with modesty and lead to personal epiphany: When Kakua, one of the early Zen wise men in Japan in the ninth century, returned from a journey through distant lands, the emperor asked him to come and tell him everything, what he had experienced on his journey. Kakua bowed deeply, remained silent for a long time, took the small bamboo flute, which he always carried with him, out of his pocket and played a single note, remained silent, bowed even deeper than before and left. The emperor was at a loss. But he kept this one note in his heart, and—it is reported—in old age he found enlightenment (Berendt 1998).

“Twelve-Tone Music Reloaded”: 12 Lessons in Rotating. . .

221

2 Designing the “Organizational Groove” These 12 principles mentioned above, which consist of the 7 key rules of rotating leadership and the 5 general lessons for teamwork from Jazz, can be extended toward designing an “organizational groove.” When one of the authors asked Yaron Gershovsky, musical conductor of the legendary American Vocal Quartet “The Manhattan Transfer,” during their latest European Tour in fall 2018, “how can you describe groove and are there “honest signals” of the audience that can be predicted?”, the internationally renowned pianist described it as follows: Playing for an audience of 10 people or a thousand people has very many similarities. I view an audience as one being, may it be large or small. They all feel and respond as one to what they hear. They all can sense the player’s feelings. They sense the inspiration and they also sense the nervousness and insecurity if it exists. When soloing, you always look for this instance when you know that you ‘got’ them. You grabbed their attention and they all are with you for the ride. It’s a great feeling! Like a ‘Runner’s High’ . . . you could call it the ‘Player’s High’ . . . This last European tour took me from Slovakia to Finland, from Sweden to Switzerland . . . the audience’s behavior was similar. It sensed my feelings and responded to them. Great experience! (Yaron Gershovsky, quoted on 18 November 2018)

Interestingly, Gershovsky, who has been in the Jazz business for almost 50 years, doesn’t mention any cultural difference of the target audience between so-called linear-active cultures such as Germany, Sweden, and Switzerland, compared to a rather “reactive” (¼ introverted) population like in Finland or a so-called multiactive (¼ extraverted) culture as we find it in Slovakia (Lewis 2018). If we follow Gershovsky’s argument, we can assume that for him as a real “Jazz Pro” neither the size of the public nor the cultural location matters, as long as Duke Ellington’s quote is still valid: “It don’t mean a thing (if it ain’t got that Swing)!” Imagine how ordinary listeners—and not only “Jazz aficionados”—can recognize Jazz standards by just focusing on different parameters: well-known melody-based tunes like “Girl from Ipanema” by Antonio Carlos Jobim differ a lot from standards with complex harmonic structures such as John Coltrane’s “Countdown” or Dave Brubeck’s “Take 5,” which achieved iconographic status with its 5/4 beat. The main issue is that the formal development of Jazz never can be predicted, as Thelonious Monk wisely mentioned a long time ago: “I don’t know where it’s going. Maybe it’s going to hell. You can’t make anything go anywhere. It just happens” (Chinen 2018). What about the idea of transferring this Jazz metaphor into an agile organizational design structure: Can this make sense at all, if Jazz itself is not predictable? In order to delve deeper into a corporate’s DNA, we propose to adapt Ed Schein’s “Culture Model” (Schein and Schein 2017, s. also Fatzer et al. 2019) to Jazz by aligning the three main musical components “melody,” “harmony,” and “beat” to his concept of surface and deep structure. Assuming that a Jazz tune always consists of these three fundamental layers, we suggest an agile organizational design that corresponds with the basic structures of a Jazz standard (Fig. 1).

222

D. C. Schmid and P. A. Gloor

Fig. 1 Schmid, D.C. Gloor, P. “Twelve-Tone Music Reloaded” (Schmid and Gloor 2018)

The screening of an organization can be tackled in an “outside-in” mode, such as moving from “surface” to “deep structure” in Ed Schein’s approach. He differentiates between “artifacts and symbols” (¼ “melody,” what can be heard at the drop of a hat), “espoused values” (¼ “harmony,” which requires careful listening to structures), and “assumptions” (¼ “beat,” the basic clock pulse of an organization). In the ideal world of Jazz, these three factors come together and create this “flow” with the audience that Yaron Gershovsky has described above. What about the economic reality of organizations and their stakeholders? As Shannon and Weaver have described in their model of communication in the late 1940s, a communication is successful if the sent message is identical to the received one, whereas the process of signal transmission is considered to be particularly susceptible to interference (Shannon and Weaver 1963). Listening carefully to all the signals which are audible means that these can be possibly distorted by noise. If this is the case, then the key messages of a corporation (such as its USP (“unique selling propositions”), EVPs (“employee value proposition”), information relating to investor relations, etc.) can’t be fully recognized by the intended target audience consisting of all stakeholders (clients, employees, shareholders, etc.). We therefore propose that understanding the “DNA code” of an organization requires to identify all existing honest signals within a company. By using the “happimeter sensing system” which allows users to predict their mood through body signals with smartwatches (Gloor et al. 2018), we aim to initiate a debate about combining “digital” and “analogue” signals to fully detect a corporation’s DNA. Aligning the three Jazz layers “melody” (¼ communication content), “harmony” (¼ structure and processes), and “beat” (lifecycle of a company) can be key

“Twelve-Tone Music Reloaded”: 12 Lessons in Rotating. . .

223

Fig. 2 Schmid, D.C. Gloor, P. “Twelve-Tone Music Reloaded” (Schmid and Gloor 2018)

to analyzing the different patterns within various organizations. In order to create a “Corporate Groove,” our goal is to take the next steps toward a theory of “organizational flow” to measure intrinsic organizational “groove.” This rollout may be supported by the happimeter technology that can be deployed in an entire global corporation (Fig. 2).

3 Discussion While COINs form when people from many different backgrounds, countries, and cultures get together to innovate toward a shared goal, Jazz musicians are special members of COINs, as they already share the same context to a large extent and thus do not have to spend long hours to build a shared language and understanding. Also, COINs frequently collaborate over long distance using the Internet to form virtual teams, while Jazz bands normally share the same stage. Therefore, Jazz bands are “elite” COINs, sharing a privileged environment that “ordinary” COINs do not have. They thus are idealized role models and exemplars illustrating the key tenets that COINs using virtual collaboration techniques should look for to gain inspiration and deep insights. As we can see, the future of virtual musical collaboration is already there. Tools such as Apple’s “GarageBand” are creating an online ecosystem that allows musicians all over the world to improvise together, no matter where they live. Thanks to

224

D. C. Schmid and P. A. Gloor

these virtual jam sessions, cultural, organizational, and professional boundaries are becoming increasingly obsolete. The limits for this cooperation are set on the one hand by the capacity of the respective Internet access and on the other hand by the technical skills of the Jazz musicians. By using the “Jam Session” mode, they get connected by the bandleader who initializes a session, and then up to three band members can join real time. Right now, the bandleader can chose between taking exclusive control of playback and recording or allow the other members to share control. Probably, the next step will be a rotating leadership mode, in order to fulfill one of the main COIN principles. The bet is on. By combining the seven key principles of rotating leadership, shown by the two ragtime pianists Tom Brier and Adam Swanson, with the five key principles of organizational development shown by personal experience of international Jazz authorities, we get 12 success factors that can be effectively used for (re)designing the organization. This set of “twelve-tone music” is—on the contrary to its atonal roots and strict principles—completely free in its combination. Our goal is to take the next steps toward a theory of “Organizational Flow” in order to measure intrinsic organizational “groove” with sociometric badges (Gloor et al. 2013) and other devices that measure human emotions such as the new happimeter technology (Gloor et al. 2018). Thus, we get the perfect quantitative conditions to start measuring the growth of a company. In other words, “organizational flow” is where the groove of an organization starts becoming effective!

References Adorno, T. W. (1936). Über Jazz. Zeitschrift für Sozialforschung, 5, 235–259. (in German). Barrett, F. (2012). Yes to the mess. Surprising leadership lessons from Jazz. Boston: Harvard Business Review Press. Berendt, J. E. (1998). Ich höre, also bin ich. In T. Vogel (Ed.), Über das Hören: Einem Phänomen auf der Spur (pp. 69–70). Tübingen: Attempto. (in German). Berendt, J. E., & Adorno, T. W. (1953). Für und wider den Jazz. Merkur, 7(67), 887–893. (in German). Burow, O. (2004). How organizations become creative fields: The Jazz band model of leadership and the role of improvisation. Zeitschrift für Supervision, 2, 1–21. (in German). Butterfield, M. W. (2011). Why do Jazz musicians swing their eighth notes? Music Theory Spectrum, 33(1), 3–26. Chinen, N. (2018). Playing changes. Jazz for the new century. New York: Pantheon Books. De Pree, M. (2008). Leadership Jazz (2nd ed.). New York: Doubleday. Fatzer, G., Van Maanen, J., Schmid, D. C., & Weber, W. (Eds.). (2019). Edgar H. Schein. The spirit of inquiry. Innsbruck: Innsbruck University Press. Gloor, P., Colladon, A. F., Grippa, F., Budner, P., & Eirich, J. (2018). Aristotle said “happiness is a state of activity”—Predicting mood through body sensing with Smartwatches. Journal of Systems Science and Systems Engineering, 27(5), 586–612. Gloor, P., Oster, D., & Fischbach, K. (2013). JazzFlow – Analyzing “group flow” among jazz musicians through “honest signals”. Artificial Intelligence, 27(1), 37–43. Hodson, R. (2007). Interaction, improvisation, and interplay in Jazz. New York/London: Routledge.

“Twelve-Tone Music Reloaded”: 12 Lessons in Rotating. . .

225

Johansen, R. (2012). Leaders make the future. Ten new leadership skills for an uncertain world (2nd ed.). Oakland: Berrett-Koehler. Kamoche, K., Pina e Cunha, M., & Vieira da Cunha, J. (2003). Towards a theory of organizational improvisation: Looking beyond the Jazz metaphor. Journal of Management Studies, 40(8), 0022–2380. Kidane, Y., & Gloor, P. (2007). Correlating temporal communication patterns of the eclipse open source community with performance and creativity. Computational & Mathematical Organization Theory, 13(1), 17–27. Lewin, K. (1997). Resolving social conflicts, and field theory in social science. Washington, DC: American Psychological Association. (Reprint. Original Harper and Row, New York 1948). Lewis, R. (2018). When cultures collide. Leading across cultures (4th ed.). London/Boston: Nicholas Brealey International. Nemoto, K., Gloor, P., & Laubacher, R. (2011). Social capital increases efficiency of collaboration among Wikipedia editors. In ACM Hypertext 2011: 22nd ACM Conference on Hypertext and Hypermedia, Eindhoven, NL, June 6–9, 2011. Scharmer, O. (2016). Theory U: Learning from the future as it emerges. The social technology of presencing. Oakland: Berrett-Koehler. Scharmer, O. (2018). The essentials of theory U: Core principles and applications. Oakland: Berrett-Koehler. Schein, E., & Schein, P. (2017). Organizational culture and leadership (5th ed.). Hoboken: Wiley. Schein, E., & Schein, P. (2018). Humble leadership: The power of relationships, openness, and trust. Oakland: Berrett-Koehler. Schmid, D. C., & Gloor, P. (2018). “Twelve-tone music reloaded”: 12 lessons in rotating leadership and organizational development from Jazz. In Proceedings of the 8th international conference on collaborative innovation networks (COINs), Suzhou, China, September 26–28, 2018. Senge, P. (2006). The fifth discipline: The art and practice of the learning organization (6th ed.). New York: Currency Doubleday. Shannon, C., & Weaver, W. (1963). The mathematical theory of communication. Champaign: University of Illinois Press. Weick, K. E. (1999). The aesthetic of imperfection in organizations. In M. P. Cunha & C. A. Marques (Eds.), Readings in organization science (pp. 541–563). Lisbon: ISPA.